Parameter control method of breast pump and application

Abstract

The present disclosure provides a parameter control method and application of a breast pump. It relates to the field of mother and baby technology. The parameter control method of the breast pump comprises: detecting a milk cluster lactation physiological event, determining whether a milk arrival event or a milk cluster event occurs; capturing the current operating parameters of the breast pump when the milk arrival event or the milk cluster event occurs, generating an event snapshot with added label data and containing a structured parameter vector, and the structured parameter vector at least includes a vacuum degree parameter, a breast pumping frequency parameter, and a suction air duty cycle parameter; persistently storing the event snapshot to a snapshot storage set; in response to a breast pump start request, matching a target event snapshot from the snapshot storage set according to a preset screening strategy; loading the parameter vector of the target event snapshot, and obtaining the current operating parameters of the breast pump after version mapping and hardware adaptation adjustment. The present disclosure continuously optimizes the parameter matching effect with user use, adapts to the changing needs of users in different lactation stages, and improves the intelligent level of the breast pump and the user experience.

Description

Technical Field

[0001] This manual relates to the field of maternal and infant products technology, and in particular to a parameter control method and application for a breast pump. Background Technology

[0002] The parameter control method for breast pumps is a technique that simulates the natural sucking rhythm of an infant by adjusting parameters such as suction power, frequency, and mode, thereby improving breastfeeding efficiency, comfort, and milk production. This method aims to provide breastfeeding mothers with a more scientific and personalized breastfeeding experience, while better meeting the feeding needs of both mother and baby.

[0003] Existing breast pump parameter control methods often suffer from insufficient precision in practical applications, lack of personalized settings, and poor compatibility with the baby's natural sucking patterns. This leads to poor milk expression, a poor mother's experience, and may even cause discomfort such as breast pain. Furthermore, traditional breast pump parameter control methods are not adaptable to mothers at different physiological stages and with varying breast conditions, failing to meet diverse needs. A more intelligent and precise parameter control method is urgently needed to address these shortcomings. However, different mothers have different breast sensitivity and milk production levels, and the needs of the same mother may change at different breastfeeding stages. Traditional breast pump parameter control methods struggle to accurately adapt to these individual needs, easily leading to problems such as incomplete milk expression or overstimulation.

[0004] Therefore, there is an urgent need for an intelligent control method that can dynamically adjust parameters according to the actual situation of the mother and baby in order to improve the scientific nature and comfort of breast pumping. Summary of the Invention

[0005] This manual provides a parameter control method for a breast pump, which can solve the problems existing in related technologies.

[0006] This disclosure provides a method for controlling the parameters of a breast pump, including: Detect milk ejection physiological events to determine whether a milk-inducing event or a milk ejection event has occurred. When a milk initiation or let-down event occurs, capture the current operating parameters of the breast pump and generate an event snapshot with added tagged data and a structured parameter vector. The structured parameter vector includes at least the vacuum parameter, the milk pumping frequency parameter, and the suction duty cycle parameter. Persist the event snapshot to the snapshot storage collection; In response to the breast pump start request, the target event snapshot is obtained by matching from the snapshot storage set according to the preset filtering strategy; The parameter vector of the target event snapshot is loaded, and after version mapping and hardware adaptation adjustments, the current operating parameters of the breast pump are obtained.

[0007] The breast pump parameter control method disclosed herein automatically identifies milk let-down and retains users' personalized and effective parameters, achieving automatic adaptation of breast pumping parameters based on users' own historical usage experience. This solves the problem that traditional breast pumps use factory-standard parameters that cannot adapt to the different lactation rhythms and breast sensitivity differences of different users, and also eliminates the tedious process of users manually adjusting parameters. It can balance breast pumping comfort and efficiency while matching the natural lactation pattern of the human body. At the same time, the persistent storage mechanism can continuously optimize the parameter matching effect as users use it, adapting to the changing needs of users at different lactation stages, thus improving the intelligence level of the breast pump and the user experience.

[0008] In some examples, detecting the physiological events of milk let-down (milk ejection reflex) includes: The negative pressure value of the breast pump is continuously collected by a pressure sensor; Calculate the change in negative pressure per unit time; When the change in negative pressure exceeds a preset threshold, a milk inflow event or a milk ejection event is determined to have occurred.

[0009] The above steps can continuously collect the negative pressure value of the breast pump by pressure sensor, calculate the change in negative pressure per unit time, and determine the milk initiation event or milk let-down event by preset threshold. There is no need to add an additional dedicated physiological detection module. The detection can be completed by relying only on the pressure sensing structure configured in the breast pump itself. It does not require changing the overall structure of the breast pump, does not increase the hardware manufacturing cost of the breast pump, and is compatible with the upgrade and transformation needs of most existing electric breast pumps, making the threshold for implementation and promotion lower.

[0010] In some examples, detecting the physiological events of milk let-down (milk ejection reflex) includes: It receives milk ejection signal signals manually triggered by the user through physical buttons on the breast pump or by operating the terminal software, and determines whether a milk initiation event or a milk ejection event has occurred.

[0011] The above-mentioned methods of manually triggering the milk arrival event or milk burst event can adapt to the usage habits and personalized scenario needs of different users.

[0012] In some examples, the tag data for the event snapshot includes at least a breastfeeding side identifier and a collection period tag; The breast pump side indicator is automatically identified and determined by sensors, or can be manually selected and determined by the user; The collection period labels are automatically generated based on the current time.

[0013] By setting up a label for the breast pumping side, it is possible to accurately distinguish the milk pumping data of different breasts, helping users and medical staff to clearly understand the milk production status of each breast, avoiding data confusion between different sides, providing an accurate basis for judging whether the milk production of both breasts is balanced, helping users to adjust their pumping habits in a timely manner, providing targeted breast care, and reducing the risk of unilateral breast stasis and inflammation.

[0014] In some examples, the steps of matching target event snapshots from the snapshot storage collection according to a preset filtering strategy include: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than the preset number threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots that meet the search criteria in the snapshot storage set is greater than or equal to a preset threshold, a weighted centroid filtering strategy is used to generate the target event snapshot.

[0015] The above steps integrate effective information from multiple snapshots, improving the accuracy of the results. The weighted centroid screening strategy integrates the feature information of all snapshots that meet the search criteria, assigns weights corresponding to the matching degree to different snapshots, and calculates the feature centroids to generate the target snapshot, rather than selecting only a single snapshot. This fully integrates the effective information from multiple candidate snapshots, offsetting the impact of errors and data anomalies that may exist in a single snapshot on the results. The final target snapshot is closer to the true characteristics of the event, adapting to the need to fully utilize data in large-sample scenarios.

[0016] In some examples, the preset filtering strategies include: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than the preset number threshold, a time-series filtering strategy is adopted to directly select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots in the snapshot storage set that meet the search criteria is greater than or equal to a preset threshold, the system automatically switches to a weighted centroid filtering strategy. For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot. For all snapshots that meet the criteria, each parameter is summed with weights and then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is then constructed using the weighted centroid parameter vector.

[0017] This disclosure, through the aforementioned hierarchical filtering strategy, allows for flexible selection of the appropriate filtering method based on the number of event snapshots that meet the search criteria, balancing retrieval efficiency with the quality of the generated target event snapshots. The automatic switching mechanism between the two filtering strategies requires no manual intervention and can adapt to filtering needs under different data volumes. It outputs high-quality target results in both data-sparse and data-rich scenarios, improving the environmental adaptability and overall operational efficiency of the entire event snapshot retrieval system.

[0018] In some examples, weighted centroid filtering strategies include: For each event snapshot that meets the search criteria, the weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot; For all snapshots that meet the criteria, each parameter is weighted and summed, then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is then constructed using this weighted centroid parameter vector.

[0019] The above steps can reduce noise data interference and improve the reliability of the results. The milk yield gain parameter of different event snapshots represents the actual contribution of the snapshot to the target milk production event, and the parameter confidence level represents the credibility level of the current snapshot measurement and calculation results. Using the product of the two as the snapshot weight allows effective snapshots with higher gain and stronger credibility to occupy a higher weight in the centroid calculation, which greatly reduces the negative impact of low confidence and low gain noise snapshots on the final result, avoids abnormal data interfering with the accuracy of the target event features, and improves the robustness of the overall results.

[0020] In some examples, the preset filtering strategy also includes a user star filtering strategy, which selects the user starred event snapshot as the target event snapshot when there are event snapshots in the snapshot storage collection that meet the search criteria.

[0021] The aforementioned user star-based filtering strategy can be adapted to lactation parameter management. Specifically, by introducing a user star-based filtering strategy into the breast pump's parameter storage and retrieval system, it can accurately match the user's own lactation needs and priorities, solving the problem that general parameter retrieval and sorting logic cannot accurately identify the user's individual preferences.

[0022] In some examples, the process of loading the parameter vector of a target event snapshot includes: First, read the firmware version of the current device and the model of the currently connected pump body, and then compare them with the original firmware version and the original pump body model recorded in the target event snapshot. If there are version differences, the original parameter number is mapped to the parameter number corresponding to the current firmware version through a pre-stored version mapping table.

[0023] The above verification and matching steps can effectively solve the problem of incompatibility of old lactation event snapshot parameters caused by firmware updates and hardware parameter adjustments during the product iteration and upgrade process of breast pumps through the pre-verification and matching mechanism of breast pump firmware version and pump hardware model. This avoids failures such as snapshot loading failure and lactation event data parsing errors caused by parameter number mismatch, and greatly improves the compatibility and traceability of historical event data between different version iterations of products.

[0024] In some examples, the steps after obtaining the mapped parameters also include: The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump hardware to obtain a parameter vector adapted to the current hardware, which serves as the current operating parameters of the breast pump.

[0025] By cropping and scaling the snapshot parameters to fit the current pump hardware, it can be ensured that the output operating parameters always fit within the parameter boundary range of the target hardware, avoiding problems such as breast pump malfunction and component overload damage caused by parameters exceeding the limit, and effectively extending the service life of the breast pump hardware.

[0026] In some examples, the structured parameter vector of the event snapshot includes at least: vacuum level parameter, milk suction frequency parameter, and suction duty cycle parameter.

[0027] By integrating vacuum parameters, milk expression frequency parameters, and suction duty cycle parameters into a structured parameter vector to construct an event snapshot, the above method can accurately capture the actual milk expression needs of different lactation stages and individuals. This solves the problem that the fixed parameter mode of traditional breast pumps cannot adapt to individual physiological differences in lactation, and greatly improves the comfort and adaptability of using the breast pump.

[0028] In some examples, the event snapshot includes at least one of the following tag data: The data includes the following information: milk pumping side identifier, data collection period label, firmware version number at the time of data collection, pump model identifier at the time of data collection, user star mark, milk volume gain parameter at the time of data collection, and parameter confidence level parameter.

[0029] The above-mentioned data can be structured and labeled in all dimensions of the breast pump's working process by configuring one or more of the following labels for the event snapshot: milk pump side identifier, collection period label, firmware version number at collection time, pump model identifier at collection time, user star mark, milk volume gain parameter at collection time, and parameter confidence parameter. This provides a clearly labeled sample dataset for subsequent parameter optimization model training.

[0030] In some examples, before the step of responding to a breast pump start request and matching the target event snapshot from the snapshot storage collection according to a preset filtering strategy, a side-by-side and time-segmented bucket retrieval step is also included: Obtain the milk pumping side identifier and the time period to which the current milk pumping request belongs, and match event snapshots from the snapshot storage set that have the same milk pumping side identifier and the same time period label to participate in the filtering.

[0031] The above-described side- and time-segmented bucket search steps can accurately limit the search scope based on the differences in lactation patterns of different breasts. This aligns with the physiological characteristics of most users, such as the natural differences in lactation volume and lactation sensitivity between the left and right sides, and the regular fluctuations in human lactation volume throughout the day and night. It significantly reduces the computational workload of irrelevant snapshot data, improves the matching efficiency of target event snapshots, and avoids interference from historical lactation data from different sides and time periods with parameter matching results, making the matched working parameters more closely match the actual lactation status of the current breast at the current time.

[0032] In some examples, when no event snapshot matches the current breastfeeding side identifier, a separate search for each side is not triggered, and a global snapshot is used for filtering. If no event snapshot matches the time period label of the current time, time period matching is not triggered, and all snapshots of the corresponding breast pumping side are used for filtering.

[0033] The adaptive retrieval strategy based on the above branch judgment can flexibly adjust the retrieval scope according to the different data coverage of the actual use scenario of the breast pump. It retains the advantage of accurate matching by breast pump side and time period, while avoiding the problem of recommendation model failure due to lack of corresponding matching snapshots, ensuring that the breast pump parameter recommendation process can always run stably.

[0034] In some examples, cross-version compatibility mapping steps are also included: When loading the parameters of the event snapshot, read the firmware version of the current device and the model of the currently connected pump body, and compare them with the original firmware version and the original pump body model recorded in the event snapshot. If there are version differences, the original parameter number will be mapped to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

[0035] The above steps, through cross-version compatibility mapping, can achieve seamless compatibility of breast pump parameter event snapshots between different firmware versions and different pump hardware models. This avoids the problem of old parameter snapshots being unable to be loaded and used due to version iterations and hardware updates, eliminating the need for users to manually reset parameters and greatly improving the user experience after changing devices or upgrading firmware.

[0036] In some examples, parameter boundary clipping includes: If the vacuum parameter in the event snapshot is higher than the current pump body hardware's preset negative pressure safety limit, the vacuum parameter will be automatically adjusted to the negative pressure safety limit, while retaining the proportional relationship of the other parameters.

[0037] By using the above parameter boundary trimming rules, it is possible to ensure that the overall negative pressure output characteristics of the breast pump in the current lactation mode match the user's customized lactation rhythm, while avoiding the hardware from working in an overload state that exceeds the threshold for a long time. This not only prevents excessive negative pressure from causing unnecessary stimulation and damage to the user's mammary glands, but also avoids the risk of malfunctions such as accelerated aging and sealing failure of the pump motor and air circuit structure due to overpressure operation.

[0038] Some examples also include the milk array window linkage control steps: When a milk ejection event is detected to enter the milk ejection duration window, a short-term enhancement adjustment is triggered, and the current operating parameters are adjusted to the enhancement parameters. When generating an event snapshot, the running parameters are extracted from the moment after the milk flow occurs and before the enhanced adjustment takes effect. The enhanced parameters after the short-term enhanced adjustment are not recorded.

[0039] By setting up a linked control step for the milk let-down window, parameters can be adjusted specifically during the milk let-down phase to enhance milk expression efficiency. This not only matches the physiological rhythm of lactation and shortens the milk expression time, but also avoids excessive breast stimulation caused by maintaining high power parameters throughout the process, thus improving user comfort.

[0040] In some examples, a snapshot incremental update step is also included: When a user manually adjusts the running parameters in the current lactation session, a confirmation prompt is output asking whether to update the snapshot set, and the user's confirmation response is received. If the user confirms the update, the adjusted parameters will be written as a new event snapshot based on the current capacity of the snapshot storage set. When the storage capacity exceeds the limit, the event snapshot generated earliest will be replaced. If the user confirms that they will not update, the adjusted parameters will be valid for the current lactation session and will not be written to the snapshot storage set.

[0041] By introducing a snapshot-based progressive update process, it is possible to accurately adapt to the personalized parameter adjustment needs of different users in the same lactation session, respecting the user's autonomy while continuously optimizing the data source for intelligent parameter recommendations.

[0042] In some examples, when the user confirms the update, it is determined whether the difference between the currently manually adjusted parameters and the target parameters in the existing snapshot exceeds a preset difference threshold. If it does not exceed the preset difference threshold, the weight parameters of the corresponding existing snapshot are updated directly without adding a new snapshot entry.

[0043] By determining whether the difference between the manually adjusted parameters and the existing snapshot target parameters exceeds a preset difference threshold, the existing snapshot weight parameters are updated directly without adding new snapshot entries only if the difference does not exceed the threshold.

[0044] In some examples, snapshot processing steps are also included in a special security mode: When entering the special security mode, security audits are performed on all event snapshot parameters to be loaded in the snapshot storage set. Parameters that exceed the preset parameter boundaries of the special security mode are forcibly pruned to within the boundary values. An audit flag is added to the pruned event snapshots, and the pruned parameters are loaded without discarding the original snapshot data.

[0045] The above-described steps of performing security auditing, boundary clipping, and adding audit flags on all event snapshot parameters to be loaded in a special security mode can achieve the following beneficial effects: Ensuring safe operation of the breast pump and mitigating safety risks: For special scenarios such as abnormal battery voltage, excessive motor load, and excessive pump pressure, the excessive parameters are forcibly reduced to within safe limits. This avoids problems such as motor overload, pressure runaway, and circuit overcurrent caused by parameters exceeding the range, and eliminates the need to directly trigger shutdown and lockout. It allows the breast pump to maintain continuous operation within safe power and pressure ranges, ensuring user safety and preventing sudden interruptions in pumping from affecting the user experience.

[0046] In some examples, the special safety mode is a nipple hypersensitivity mode or a premature infant breastfeeding mode, and the upper limit of negative pressure of the preset parameter boundary of the special safety mode is lower than the upper limit of negative pressure of the normal mode.

[0047] By setting the preset negative pressure limit of the special safety mode to be lower than that of the normal mode, targeted safety protection can be provided for the special needs of users with highly sensitive nipples and for breastfeeding scenarios of premature infants.

[0048] In some examples, a cold start process is triggered when no event snapshot corresponding to the current search criteria exists in the snapshot storage collection. The cold start process includes: Obtain the group prior parameter vector of users with the same breast pump model and nipple size from the cloud, and use the group prior parameter vector as the initial recommendation parameters. After each new individual event snapshot is generated, the fusion weight of the group prior parameters is reduced; When the number of individual event snapshots reaches the preset cold start threshold, the fusion weight of the group prior parameters is adjusted to zero, and the filtering results of individual snapshots are used.

[0049] The above-described cold start process effectively addresses the pain point of inaccurate parameter recommendations and the need for users to repeatedly manually adjust the settings when using a breast pump, due to the lack of individual historical usage data.

[0050] Initial recommendations are made based on prior parameters of the same model of breast pump and the same nipple size group. The initial recommended parameters are in line with the general usage habits of this group. Compared with random initialization parameters or fixed default parameters, the initial comfort and lactation efficiency are better, which greatly reduces the frequency and operational burden of new users adjusting parameters in the early stage and improves the experience of first-time users.

[0051] In some examples, the group prior parameter vector is the cluster centroid parameter obtained by clustering valid event snapshots of users of the same model and size in the cloud.

[0052] The aforementioned group prior parameter vector, obtained by clustering effective event snapshots of users of the same model and size, serves as the initial benchmark for breast pump parameter control. This approach aligns with the general lactation patterns and usage preferences of the target user group, avoiding the lengthy trial-and-error process of adapting a brand-new device to individual users from scratch. It significantly shortens the waiting time for users to obtain suitable parameters and reduces the probability of breastfeeding pain and low lactation efficiency caused by parameter mismatch during the initial use of new users.

[0053] On the other hand, this disclosure also provides a breast pump control system for executing the above-described parameter control method for a breast pump. The breast pump control system includes a physiological event detection module, a snapshot generation and storage module, a snapshot filtering and matching module, and a parameter loading module. The physiological event detection module is used to continuously collect the negative pressure value of the breast pump through a pressure sensor, calculate the change in negative pressure per unit time, and determine the occurrence of a milk initiation event or milk let-down event when the change in negative pressure exceeds a preset threshold, or receive a milk let-down marker signal manually triggered by the user. The snapshot generation and storage module is electrically connected to the physiological event detection module. It is used to capture the current operating parameters at the moment a physiological event occurs, generate an event snapshot containing structured parameter vectors and label data, and persistently store the event snapshot in the snapshot storage collection. The snapshot filtering and matching module is electrically connected to the snapshot generation and storage module. Upon receiving a startup request, it matches the target event snapshot from the snapshot storage set according to a preset filtering strategy. When the number of event snapshots that meet the search criteria in the snapshot storage set is less than a preset threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots that meet the search criteria in the snapshot storage set is greater than or equal to a preset threshold, a weighted centroid filtering strategy is used to construct the target event snapshot. The parameter loading module is used to load the parameters of the target event snapshot as the current running parameters. During loading, it first reads the firmware version of the current device and the model of the currently connected pump body hardware, compares the original firmware version and the original pump body model recorded in the event snapshot, and if there is a version difference, it maps the original parameter number to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. Then, it trims and scales the snapshot parameters according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

[0054] In the above structure, the physiological event detection module combines automatic detection and manual triggering for milk ejection reflex recognition, adapting to different user habits and recognition needs in different scenarios: it collects negative pressure data through pressure sensors and determines milk ejection reflexes based on changes per unit time, automatically capturing milk ejection or milk initiation events without user manual operation, reducing the user's operational burden during breast pumping. At the same time, the algorithm logic is simple, has low requirements for device computing power, and does not increase the hardware cost of the breast pump. The design that supports users to manually trigger milk ejection reflex marker signals can deal with special cases of missed or false detections by automatic detection. For example, when some users' milk ejection reflexes arrive, the negative pressure change does not meet the preset threshold, or external interference during breast pumping causes abnormal fluctuations in negative pressure, leading to false detections. Manual marking can correct the deviation of automatic detection, greatly improving the accuracy and flexibility of milk ejection or milk initiation event recognition, and providing a reliable triggering basis for generating accurate event snapshots in the future.

[0055] In some examples, users manually trigger the let-down signal by pressing a physical button on the breast pump or by operating the terminal software.

[0056] The solution that triggers the milk let-down signal via physical buttons requires no additional modification to the breast pump's hardware structure. It can be achieved using the operation buttons already configured on the breast pump itself, resulting in low modification costs. It is compatible with most existing mass-produced breast pump products, and the operation is intuitive and convenient for users. It can quickly complete the marking without the need for an additional smart terminal connection, making it suitable for breast pumping scenarios without network access or mobile devices.

[0057] In some examples, weighted centroid filtering strategies include: For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk volume gain parameter and the parameter confidence parameter of that snapshot. The weighted sum of each parameter of all snapshots that meet the criteria is then divided by the sum of all weights to obtain the weighted centroid parameter vector. The target event snapshot is constructed using the weighted centroid parameter vector.

[0058] By calculating the weight of a single event snapshot through the product of the milk yield gain parameter and the parameter confidence parameter, we can simultaneously take into account the actual milk yield improvement effect during the milk pumping process and the reliability of the current parameter combination. This avoids the bias caused by selecting parameters based solely on milk yield or model confidence. It will not ignore high-potential effective parameter combinations under low milk yield gain, nor will it exclude personalized parameters with low confidence that are actually suitable for the current milk yield status, thus improving the scientific nature of parameter selection.

[0059] In some examples, a program module for performing the breast pump control method described above is also included.

[0060] By integrating the program module into the control unit of the breast pump, modular deployment and iterative updates of the control logic can be achieved.

[0061] In some examples, a cloud server is also included, which stores snapshot data of the group of users and is used to send the group's prior parameter vectors corresponding to the model and nipple size to the breast pump upon request. The group prior parameter vector is the cluster centroid parameter obtained by clustering effective event snapshots of users of the same model and size in the cloud.

[0062] By storing snapshot data of aggregated users on cloud servers, clustering calculations can be performed based on massive amounts of real usage data from users with the same breast pump model and nipple size. This yields a group prior parameter vector that better reflects the actual usage needs of most users. Compared to individual users having to figure out and adjust parameters from scratch, users can directly download initial parameters that match their own situation after unpacking the device. This significantly shortens the parameter adaptation cycle for novice users, reduces the operational costs of manual adjustments, avoids problems such as breast pumping pain and low lactation efficiency caused by unreasonable initial parameter settings, and improves the initial user experience.

[0063] On the other hand, this disclosure also provides a breast pump, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the parameter control method of the breast pump described above.

[0064] The aforementioned breast pump, through its preset parameter control logic, compared to traditional breast pumps that use fixed lactation modes, fixed vacuum pressure levels, and fixed pumping rhythms, can automatically and dynamically adapt pressure and sucking frequency according to the individual user's mammary gland condition, lactation pattern, and pain tolerance. This avoids problems such as nipple and areola damage and breast engorgement caused by excessive pressure, and also avoids incomplete lactation and low milk expression efficiency caused by insufficient pressure or mismatched rhythm, effectively improving the comfort and thoroughness of the milk expression process.

[0065] On the other hand, this disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the parameter control method for the breast pump described above.

[0066] By solidifying the parameter control logic into an independently callable computer program and storing it in a readable storage medium, the breast pump parameter control method can be quickly ported and deployed across devices. Whether it is used to upgrade the firmware of existing breast pump products or integrated into newly developed smart breast pump hardware, there is no need to redevelop the core control logic, which greatly reduces the R&D iteration cost of smart breast pumps and shortens the product launch cycle. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 A flowchart illustrating a parameter control method for a breast pump according to an embodiment of the present disclosure is shown.

[0069] Figure 2 A schematic diagram of multi-source fusion confidence score calculation is shown in a parameter control method for a breast pump provided according to an embodiment of the present disclosure.

[0070] Figure 3 The diagram illustrates the side-by-side and time-by-time bucketing retrieval logic block diagram in a parameter control method for a breast pump provided according to an embodiment of the present disclosure.

[0071] Figure 4 A schematic diagram of the system hardware connection corresponding to a parameter control method for a breast pump provided according to an embodiment of the present disclosure is shown.

[0072] Figure 5 Another flowchart illustrating a parameter control method for a breast pump according to an embodiment of the present disclosure is shown.

[0073] Figure 6 A schematic diagram of a snapshot data structure field in a parameter control method for a breast pump provided according to an embodiment of the present disclosure is shown.

[0074] Figure 7 A timing curve of smooth transition of parameters for the next session is shown in a parameter control method for a breast pump provided according to an embodiment of the present disclosure.

[0075] Figure 8A schematic diagram of an event detection state machine is shown in a parameter control method for a breast pump according to an embodiment of the present disclosure. Detailed Implementation

[0076] The following description provides specific application scenarios and requirements for this specification, intended to enable those skilled in the art to make and use the contents of this specification. Various partial modifications to the disclosed embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments and applications without departing from the spirit and scope of this specification. Therefore, this specification is not limited to the embodiments shown, but rather to the widest scope consistent with the claims.

[0077] The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not restrictive. For example, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. When used in this specification, the terms “comprising,” “including,” and / or “containing” mean that the associated feature, integer, step, operation, element, and / or component is present, but do not preclude the presence of one or more other features, integers, steps, operations, elements, components, and / or groups, or that other features, integers, steps, operations, elements, components, and / or groups may be added to the system / method.

[0078] Considering the following description, these and other features of this specification, as well as the operation and function of the related components of the structure, and the economy of assembly and manufacture of the parts, can be significantly improved. All of these form part of this specification with reference to the accompanying drawings. However, it should be clearly understood that the drawings are for illustrative and descriptive purposes only and are not intended to limit the scope of this specification. It should also be understood that the drawings are not drawn to scale.

[0079] The flowcharts used in this specification illustrate operations implemented according to some embodiments of this specification. It should be clearly understood that the operations in the flowcharts may not be implemented in a sequential order. Instead, the operations may be implemented in reverse order or simultaneously. Furthermore, one or more additional operations may be added to the flowcharts. One or more operations may be removed from the flowcharts.

[0080] In this specification, "X includes at least one of A, B, or C" means that X includes at least A, or X includes at least B, or X includes at least C. That is, X may include only one of A, B, and C, or any combination of A, B, and C, as well as other possible content / elements. Any combination of A, B, and C can be A, B, C, AB, AC, BC, or ABC.

[0081] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect, complete or partial. For example, when describing "A is connected to B," unless explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is above B," unless explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). Furthermore, when describing "A is inside B," unless explicitly stated that A is entirely inside B, it should be understood that A can be entirely inside B or partially inside B. And so on.

[0082] The parameter control method for the breast pump disclosed herein refers to a control strategy that simulates the rhythm and strength of a baby's natural sucking by adjusting parameters such as suction strength, sucking frequency, and mode switching, thereby improving breast pumping efficiency and comfort. This method aims to provide a personalized breast pumping experience based on different mothers' mammary gland conditions, breastfeeding stages, and individual tolerance, while protecting breast health.

[0083] Existing parameter control methods for breast pumps have several shortcomings. For example, the parameters lack precision, failing to accurately simulate the subtle changes in a baby's sucking movements; they lack personalized settings, making it difficult to meet the diverse needs of different mothers; and some products have complex adjustment logics and poor ease of operation, significantly impacting the user experience. Furthermore, traditional parameter control is often designed for single breastfeeding scenarios, lacking adaptability to different breastfeeding stages and unable to effectively address complex situations such as changes in milk production and blocked milk ducts. Therefore, a new parameter control method for breast pumps is needed.

[0084] This disclosure provides a parameter control method for a breast pump. Based on ergonomics and lactation principles, this method uses intelligent algorithms to monitor data such as pressure changes and milk flow rate in real time during the pumping process, dynamically adjusting suction and frequency parameters to achieve a more precise and comfortable pumping experience. Its core lies in providing personalized parameter combinations based on individual differences, while also considering ease of operation and safety.

[0085] The core difference between this disclosed method for adaptive lactation mode of a breast pump, based on the identification of milk initiation or let-down events, and existing technologies lies in the following: During breast pumping, the system monitors milk initiation or let-down events in real time. When an event occurs and the confidence level reaches a preset threshold, the system captures the relevant parameters of the breast pump's lactation mode at that moment (rather than the entire session sequence or final state) as a structured snapshot and stores it persistently. The next time the user uses the breast pump, the system automatically reads the preferred snapshot associated with that user and, within the limits of hardware capabilities, automatically sets or smoothly adjusts the lactation mode parameters to the working point recorded in the snapshot.

[0086] To facilitate the description of the structure and operating logic of the breast pump's parameter control method, unless otherwise specified, this instruction manual is based on the breast pump's natural placement posture. The "inner side" refers to the inner side of the breast shield that directly contacts the breast, located on the side of the breast pump closest to the body. Furthermore, the terms "front," "back," "left," "right," "up," and "down" in this instruction manual are defined based on the user's normal viewing angle when using the breast pump.

[0087] As an example, Figure 1 A flowchart illustrating a parameter control method for a breast pump according to an embodiment of the present disclosure is shown.

[0088] In a first aspect, this disclosure provides a method for controlling the parameters of a breast pump, including: Step S100: Detect milk ejection physiological events and determine whether a milk-inducing event or a milk ejection event has occurred; The above steps can be used to determine whether a milk-inducing event has occurred by monitoring indicators such as changes in breast fullness and milk secretion rate, and to determine whether a milk let-down event has occurred by monitoring indicators such as milk ejection frequency and breast contraction rhythm. The above steps can prevent the breast pump from always operating in a fixed mode by capturing different key physiological nodes in the user's lactation process: when no milk initiation event is detected, it maintains a low-power standby mode adapted to the lactation preparation stage, reducing unnecessary energy consumption and effectively extending the battery life of portable breast pumps; once a milk initiation event is detected, it immediately switches to a low-stimulation and gentle sucking mode adapted to the early stage of lactation, improving the comfort of breastfeeding while providing a suitable stimulation environment for milk secretion; and when the milk let-down reflex is accurately identified, subsequent parameter adjustments are triggered, which can accurately match the rhythm changes of the human body's natural lactation, providing a trigger basis for subsequent accurate adaptation of personalized breastfeeding parameters, avoiding meaningless parameter calculations, and further reducing the computing power consumption and energy consumption of the breast pump.

[0089] Step S200: When a milk initiation event or a milk let-down event occurs, capture the current operating parameters of the breast pump and generate an event snapshot with added tagged data and a structured parameter vector. The structured parameter vector includes at least the vacuum parameter, the milk pumping frequency parameter, and the suction duty cycle parameter. The above steps standardize and organize the user-suitable breast pumping parameters when milk let-down occurs: Three core breast pumping parameters—vacuum level, frequency, and duty cycle—are structured and categorized for storage, clearly representing the effective parameter characteristics corresponding to milk let-down for different users and different lactation stages, avoiding parameter mixing and inaccessibility; tagged data provides a basis for subsequent parameter matching, facilitating the differentiation of parameters for different lactation stages and usage scenarios, improving the accuracy of subsequent matching; currently effective parameters are integrated into an event snapshot, fully preserving the user's own suitable and effective personalized breast pumping parameters, achieving data-driven retention of the user's comfortable breast pumping experience, and providing a standardized data foundation for subsequent parameter reuse.

[0090] Step S300: Persist the event snapshot to the snapshot storage collection; The above steps can retain the effective personalized parameters generated during multiple uses by the user for a long time, and the data will not be lost due to restarting the breast pump or power failure. As the number of times the user uses the pump increases, the snapshot storage set will continuously accumulate parameter data that conforms to the user's own lactation pattern, gradually enriching the user's personal parameter library. This avoids the problem that the breast pump can only use the factory general parameters every time, and solves the problem that general parameters cannot adapt to the differences in lactation of different users. It provides sufficient data support for matching parameters that are more in line with the user's current needs in the future.

[0091] Step S400: In response to the breast pump start request, the target event snapshot is obtained from the snapshot storage set according to the preset filtering strategy; The above steps change the traditional breast pump's fixed parameters that require manual adjustment by the user. It can automatically match parameters that meet the current needs based on the user's own historical usage data: the preset filtering strategy can combine the user's usual pumping time, historical milk let-down probability, lactation stage and other conditions to quickly locate the most suitable historical effective parameters. Users do not need to explore and adjust the parameters again every time they use it, which greatly reduces the threshold for using the breast pump. It is especially suitable for breastfeeding novice users, simplifies the operation process and improves the user experience.

[0092] Step S500: Load the parameter vector of the target event snapshot, and obtain the current operating parameters of the breast pump after version mapping and hardware adaptation adjustment.

[0093] The above steps enable the adaptation of historical parameters: version mapping can be compatible with the parameter rules of different firmware versions of the breast pump, avoiding parameter format incompatibility and inability to call; hardware adaptation adjustment can adapt to the current working state of the breast pump hardware, such as dealing with parameter output deviations caused by changes in battery voltage and component aging, ensuring that the final output operating parameters are consistent with the parameters that provide a comfortable experience for the user in the past, and allowing the stored historical personalized parameters to be stably reused in the current operation of the breast pump, achieving accurate reuse of personalized parameters.

[0094] The breast pump parameter control method disclosed herein automatically identifies milk let-down and retains users' personalized and effective parameters, achieving automatic adaptation of breast pumping parameters based on users' own historical usage experience. This solves the problem that traditional breast pumps use factory-standard parameters that cannot adapt to the different lactation rhythms and breast sensitivity differences of different users, and also eliminates the tedious process of users manually adjusting parameters. It can balance breast pumping comfort and efficiency while matching the natural lactation pattern of the human body. At the same time, the persistent storage mechanism can continuously optimize the parameter matching effect as users use it, adapting to the changing needs of users at different lactation stages, thus improving the intelligence level of the breast pump and the user experience.

[0095] In some examples, detecting the physiological events of milk let-down and lactation includes: continuously collecting the negative pressure value of the breast pump using a pressure sensor; calculating the change in negative pressure per unit time; and determining that a milk let-down or milk initiation event has occurred when the change in negative pressure exceeds a preset threshold.

[0096] The above steps can continuously collect the negative pressure value of the breast pump by pressure sensor, calculate the change in negative pressure per unit time, and determine the milk initiation event or milk let-down event by preset threshold. There is no need to add an additional dedicated physiological detection module. The detection can be completed by relying only on the pressure sensing structure configured in the breast pump itself. It does not require changing the overall structure of the breast pump, does not increase the hardware manufacturing cost of the breast pump, and is compatible with the upgrade and transformation needs of most existing electric breast pumps, making the threshold for implementation and promotion lower.

[0097] Compared to existing milk ejection detection methods that rely on user subjective recording, breast electrical signal detection, and image recognition detection, this detection method uses the changes in negative pressure during the actual operation of the breast pump as the basis for judgment. It directly corresponds to the physiological relationship that milk ejection during the breast pumping process causes an increase in milk production, which in turn leads to a significant change in negative pressure in the breast pump cavity. The judgment logic closely matches the physical process of actual breast pumping, and the accuracy and stability of the judgment results for milk delivery events or milk ejection events are higher. It is not affected by factors such as individual differences in user perception, skin impedance interference, or interference from the shooting environment.

[0098] Based on the accurate milk let-down determination results, the breast pump can automatically match the working parameters such as the negative pressure mode and massage frequency required during the milk let-down stage, realizing automatic switching of the milk pumping mode during the pumping process without the need for manual adjustment by the user. This can not only improve the comfort of milk pumping, but also increase the total milk yield per pumping session, thus optimizing the user experience of the electric breast pump.

[0099] The data processing flow of the above detection method is simple. It only requires calculation of the change in negative pressure data collected by the pressure sensor and threshold comparison to complete the judgment. It has low requirements for the computing performance of the breast pump terminal, does not increase the power consumption of the terminal device, can adapt to the usage needs of small portable battery-powered devices such as breast pumps, and will not affect the device's battery life.

[0100] Furthermore, the above-mentioned determination method can accurately and timely capture the occurrence of milk let-down. Compared with the traditional mode that relies on users' manual perception and experience to judge milk let-down, it can avoid the judgment error caused by individual perception differences. Some users with low sensitivity cannot identify weak milk let-down in time, while some sensitive users are prone to misjudging normal lactation fluctuations as milk let-down. The determination method based on negative pressure changes relies on objective data output results, which is more universal and accurate.

[0101] Once a milk let-down reflex is detected, the breast pump can automatically match the optimal negative pressure mode for that stage: during the let-down phase, the milk flow is greater and faster, so a higher negative pressure setting can extract milk more efficiently, shorten pumping time, and eliminate the need for manual adjustments, thus improving the pumping experience. Simultaneously, accurate let-down reflex detection allows the breast pump to maintain a lower, more comfortable negative pressure during non-let-down phases, reducing pressure damage to the nipple and areola tissue from continuous high pressure, lowering the risk of nipple pain and cracking during pumping, and improving the safety and comfort of pumping.

[0102] In addition, by continuously collecting negative pressure data and identifying milk let-down reflexes, the system can accumulate statistics on the number of milk let-down reflexes during a single breastfeeding session, the duration of each reflex, and the lactation patterns at different times. This helps users clearly understand their own lactation characteristics. For breastfeeding mothers with insufficient milk production who need to adjust their breastfeeding plans, the system can provide data-driven adjustment guidelines. It can also provide objective references for lactation consultants and obstetric and gynecological medical staff to assess the lactation function of breastfeeding women and assist in the personalized development of breastfeeding guidance programs.

[0103] In some examples, detecting milk ejection physiological events includes receiving a milk ejection flag signal manually triggered by the user via a physical button on the breast pump or by operating the terminal software, and determining whether a milk initiation event or a milk ejection event has occurred.

[0104] The above-mentioned methods of manually triggering the milk arrival event or milk burst event can adapt to the usage habits and personalized scenario needs of different users.

[0105] For breastfeeding users who are able to clearly perceive their own physiological changes in lactation and are familiar with their own lactation patterns, manual marking is more in line with their own real feelings than automatic recognition by algorithms. It can effectively avoid the recognition error caused by individual differences in lactation and fluctuations in the use of breast pumps, and significantly improve the accuracy of statistics on milk initiation events or let-down events. The above method has a simple implementation logic, does not require the addition of complex sensor modules to the breast pump hardware, and does not require training of high-precision recognition algorithms. It can reduce the R&D cost and hardware power consumption of breast pump products. At the same time, it has a low user threshold, and breastfeeding users of all ages can quickly get started and complete the operation.

[0106] Manually labeled milk let-down data can also be used as labeled samples for algorithm training, which can help optimize the recognition accuracy of the automatic milk let-down recognition model and provide high-quality real data support for subsequent iterations and upgrades to a more intelligent automatic recognition function, thereby further improving the overall lactation monitoring function experience of the product.

[0107] In some examples, the event snapshot's tagged data includes at least a pumping side identifier and a collection period identifier. By setting the pumping side identifier, it is possible to accurately distinguish the pumping-related data of different breasts of the user, helping users and medical staff to clearly understand the lactation status of each breast, avoiding data confusion between different sides, providing an accurate basis for judging whether the lactation of both breasts is balanced, helping users to adjust their pumping habits in a timely manner, providing targeted breast care, and reducing the risk of unilateral breast stasis and inflammation.

[0108] The breast pump side label can be automatically identified by sensors or manually selected by the user. The automatic identification of the breast pump side eliminates the need for manual operation, recording the label automatically during the pumping process. This lowers the barrier to entry for users, making it particularly suitable for scenarios where both hands are needed to operate the pump and manual settings are inconvenient, thus improving ease of use. Supporting manual selection of the breast pump side allows for adaptation to special scenarios where sensor identification may be inaccurate. For example, if a user needs to pump from both sides simultaneously, or if the wearing position affects sensor judgment, the user can manually correct the breast pump side label, ensuring the accuracy of the label data and balancing the efficiency of automation with the flexibility of manual correction.

[0109] The data collection time period labels are automatically generated based on the current time. The above steps, through the automatically generated data collection time period labels, can intuitively present milk pumping and lactation data for different time periods. Users can clearly understand their lactation patterns at different times (such as morning, noon, night, or different weeks postpartum), making it convenient for users to adjust their pumping frequency and schedule according to their own lactation patterns. It also allows medical staff to provide more personalized lactation guidance based on changes in lactation data across different time periods, helping users establish a breastfeeding and pumping rhythm that better suits their individual physical condition.

[0110] In some examples, the steps of matching target event snapshots from the snapshot storage collection according to a preset filtering strategy include: Step S410: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than the preset number threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. The above steps ensure timely results by aligning with the latest business status. The later the event snapshot is generated, the closer it is to the current real business status of the system, and the better the data reflects the latest developments of the event. When the number of candidate snapshots is small, selecting the latest generated snapshot avoids using outdated snapshots that could lead to decisions and processing based on erroneous historical information. This adapts to the dynamic changes in event status within the business scenario and improves the accuracy of subsequent processing results.

[0111] It simplifies the filtering logic and reduces computational resource consumption. Time-series filtering only needs to compare the generation times of candidate snapshots, without performing complex weighted calculations, distance operations, or other processing steps. The filtering logic is simple and clear, with extremely low computational load. In scenarios where the number of candidate snapshots is small, results can be obtained quickly without introducing complex algorithms, significantly shortening the filtering process time, saving computational and storage resources, and improving overall processing efficiency.

[0112] It can adapt to edge scenarios and ensure stable output of the filtering process. When the number of snapshots that meet the search criteria is lower than the preset threshold, it is a edge scenario with few samples. Complex filtering strategies are prone to result deviation due to insufficient sample size. However, the results of time-series filtering are stable and predictable, and there will be no filtering failure due to a small number of candidate samples. It can ensure the continuous and stable output of target snapshots in this scenario and ensure the robustness of the entire process.

[0113] Step S420: When the number of event snapshots that meet the retrieval criteria in the snapshot storage set is greater than or equal to a preset number threshold, a weighted centroid filtering strategy is used to generate a target event snapshot.

[0114] The above steps integrate effective information from multiple snapshots, improving the accuracy of the results. The weighted centroid screening strategy integrates the feature information of all snapshots that meet the search criteria, assigns weights corresponding to the matching degree to different snapshots, and calculates the feature centroids to generate the target snapshot, rather than selecting only a single snapshot. This fully integrates the effective information from multiple candidate snapshots, offsetting the impact of errors and data anomalies that may exist in a single snapshot on the results. The final target snapshot is closer to the true characteristics of the event, adapting to the need to fully utilize data in large-sample scenarios.

[0115] It can also adapt to scenarios with ambiguous search conditions, improving matching flexibility. When there are many candidate snapshots, it often corresponds to scenarios where the search conditions are not clear enough and multiple snapshots partially meet the search requirements. Weighted centroid filtering can highlight the feature influence of snapshots with higher matching degree with the search conditions through weight allocation, and weaken the interference of snapshots with low matching degree. Compared with fixed rule filtering, it is more flexible, can better adapt to the needs of fuzzy search scenarios, and output results that are more in line with the search intent.

[0116] It can also improve the stability and representativeness of the results. A single candidate snapshot is easily affected by occasional data errors and temporary abnormal states at the time of generation, resulting in insufficient representativeness; while the target snapshot generated by weighted centroid is a concentrated reflection of the features of multiple candidate snapshots, which can represent the overall feature distribution of snapshots that meet the search conditions. The stability and representativeness of the results are far higher than that of a single snapshot selection, making it more suitable for scenarios where subsequent analysis and decision-making need to be based on the target snapshot, thereby improving the reliability of downstream tasks.

[0117] In some examples, the preset filtering strategies include: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than the preset number threshold, a time-series filtering strategy is adopted to directly select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots in the snapshot storage set that meet the search criteria is greater than or equal to a preset threshold, the system automatically switches to a weighted centroid filtering strategy. For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot. For all snapshots that meet the criteria, each parameter is summed with weights and then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is then constructed using the weighted centroid parameter vector.

[0118] This disclosure, through the aforementioned hierarchical filtering strategy, can flexibly select an appropriate filtering method based on the number of event snapshots that meet the search criteria, thus balancing search efficiency with the quality of the generated target event snapshots. When the number of event snapshots that meet the search criteria is small, the latest event snapshot is selected by directly using the time-series filtering strategy. This not only conforms to the general rule that newly generated data has higher reference value in most business scenarios, but also eliminates the complex weighted calculation process, which greatly reduces the computing power consumption and processing latency in the search and filtering process. It can quickly output the target event snapshot and meet the business system's requirements for low-latency response.

[0119] Once the number of event snapshots matching the search criteria reaches a preset threshold, the system automatically switches to a weighted centroid filtering strategy. The product of the milk yield gain parameter and the parameter confidence parameter is used as the weight of a single snapshot. This strategy reflects both the value of the event snapshot for guiding actual production through the milk yield gain parameter and the reliability of the event snapshot data itself through the parameter confidence parameter. The resulting weighted centroid parameter vector integrates information from a large number of valid snapshots, avoiding the random errors and data biases present in single snapshots. The accuracy, representativeness, and reliability of the target event snapshots are significantly improved, providing more accurate data support for subsequent livestock production decision analysis.

[0120] The automatic switching mechanism between the two filtering strategies requires no manual intervention and can adapt to filtering needs under different data volumes. It can output high-quality target results in both data-sparse and data-rich scenarios, improving the environmental adaptability and overall operating efficiency of the entire event snapshot retrieval system.

[0121] In some examples, weighted centroid filtering strategies include: For each event snapshot that meets the search criteria, the weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot; For all snapshots that meet the criteria, each parameter is weighted and summed, then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is then constructed using this weighted centroid parameter vector.

[0122] The above steps can reduce noise data interference and improve the reliability of the results. The milk yield gain parameter of different event snapshots represents the actual contribution of the snapshot to the target milk production event, and the parameter confidence level represents the credibility level of the current snapshot measurement and calculation results. Using the product of the two as the snapshot weight allows effective snapshots with higher gain and stronger credibility to occupy a higher weight in the centroid calculation, which greatly reduces the negative impact of low confidence and low gain noise snapshots on the final result, avoids abnormal data interfering with the accuracy of the target event features, and improves the robustness of the overall results.

[0123] The above strategy integrates effective information from multiple snapshots and refines core features. By using weighted summation and normalization, it integrates the feature information of all scattered event snapshots that meet the retrieval conditions into a weighted centroid parameter vector. This allows it to extract the core features common to multiple similar event snapshots, avoid the sample limitations of a single event snapshot, refine more representative target milk production event features, and construct target event snapshots that better reflect the actual milk production patterns of the target group.

[0124] It can also take into account both data contribution and credibility. The calculation logic is more scientific. Using only milk volume gain as a weight cannot distinguish the credibility of the data. Using only confidence as a weight cannot reflect the actual value of the snapshot to the milk production event. Using the product of the two as a weight takes into account both the actual contribution of the snapshot to the target event and its own data quality. Compared with single-dimensional weighting, it is more in line with the actual scenario of livestock milk production event analysis. The calculation logic is more in line with actual business needs, and the snapshot features of the target event obtained are more accurate.

[0125] This strategy can provide standardized, high-quality samples for subsequent analysis. The target event snapshots constructed through this strategy integrate information from multiple effective snapshots, resulting in more stable features and stronger representativeness. This can provide standardized, high-quality input samples for subsequent processing such as milk production pattern analysis, milk production performance prediction, and lactation event modeling, thereby reducing errors in subsequent analysis and improving the accuracy of downstream tasks.

[0126] Specifically, refer to Figure 2 The diagram illustrates the calculation of the multi-source fusion confidence score. Specifically, the milk flow / milk arrival event confidence quantification calculation module is the core algorithm unit for event judgment and filtering of falsely triggered snapshots in the patent. It can provide a confidence score for the event detection results, and not write low-confidence events into the snapshot to prevent misjudgment and contamination. It receives multiple raw signals from the sensor module and outputs a quantized score (Score) to the event state machine in Figure 3 for threshold judgment. Only high-scoring events will enter the "confirmation (write snapshot)" state, directly determining whether to generate an event snapshot record R.

[0127] The input layer in the attached diagram includes: 5 independent signal sources (multi-source feature input). All input data comes from physical signals collected by the sensing module, and each channel is assigned an independent weight for weighted summation: Negative pressure trough value |ΔN| (weight w1) When milk flows out, the negative pressure inside the shield will show obvious fluctuations and trough values. The greater the amplitude of the negative pressure change, the higher the probability of a true milk let-down, which is a positive bonus.

[0128] Flow change Q (weight w2) Pipeline flow / liquid level sensor detects a sudden increase in milk flow rate. The greater the jump in flow rate, the higher the confidence level and the positive bonus.

[0129] Pump current (weight w3): When milk enters the air circuit, it will change the pump load and the motor current characteristics will shift. If the current change characteristics match the milk flow, it will be positively scored.

[0130] User tagging (weight w4): When a user manually presses a button / the app clicks "Milk is here," it is considered a manual confirmation signal, which significantly increases confidence and adds positive points.

[0131] Leakage detection (weight w5, negative penalty): If the shield leaks or is not properly assembled, false negative pressure fluctuations will occur. When a leak is detected, the score w5*s_leak will be deducted to reduce the overall confidence level and avoid misjudging the leak as a milk flow.

[0132] Intermediate Core Computation Layer: Weighted Fusion Formula

[0133] wi: Fixed weighting coefficients for each sensor signal, which can be adjusted according to the product. si: Normalized feature score (0~1) for each sensor channel; Leakage penalty item: If there is a leak, the total score will be deducted directly to prevent false triggering.

[0134] Function: It integrates multi-dimensional information such as pressure, flow rate, current, manual marking, and leakage status into a single quantitative confidence score, thus solving the problem of misjudgment caused by a single sensor.

[0135] IV. Threshold Branch Decision Layer (Three-stage Hierarchical Logic) Two thresholds are set: a low threshold Tlow and a high threshold Thigh. The calculated score is then processed in three tiers. Score < T_low: Discard items with extremely low confidence levels, judged as interference noise such as vibration, slight air leakage, or human body shaking, do not enter the suspected state machine, and do not perform any snapshot collection operations.

[0136] T_low ≤ Score < T_high: Candidate / Confirmed (Entering "Suspected" State) The signal has milky array characteristics, but the credibility is insufficient. It is sent to the state machine for continuous M-frame anti-shake verification. Only if it maintains a high score for multiple consecutive frames will it enter the snapshot writing stage.

[0137] Score ≥ T_high: The snapshot is written with highly consistent features from multiple sensors, showing no obvious leakage and even manual marking. It is highly reliable and is directly identified as a valid physiological event. It enters the "confirmation (write snapshot)" state and immediately collects parameters to generate a snapshot.

[0138] This disclosure solves the problem of misjudgment caused by single-sensor detection. Existing Medela and Willow solutions rely solely on a single pressure / flow signal to determine milk flow, which is easily triggered by air leakage or body movement. This solution combines multi-signal fusion with leakage penalty, significantly reducing the probability of misidentification.

[0139] Implement snapshot quality filtering to prevent dirty data from polluting the snapshot library. Low-confidence events are directly discarded and will not be written to storage. This prevents false events from generating snapshots with incorrect parameters and ensures that only real milk array data is used in subsequent bucketing and selection.

[0140] Distinguishing between "automatic machine detection" and "user manual labeling," user manual labeling has an independent positive weight. When the sensor signal is weak but the user clearly perceives the milk flow, the score can be increased to balance automatic machine recognition and user subjective perception.

[0141] It is compatible with low-cost MCUs. Its computing power only performs linear weighted summation, without complex neural network operations. Embedded microcontrollers can perform real-time calculations without the need for high-end processors.

[0142] In some examples, the preset filtering strategy also includes a user star filtering strategy, which selects the user starred event snapshot as the target event snapshot when there are event snapshots in the snapshot storage collection that meet the search criteria.

[0143] The above user star selection strategy can be adapted to lactation parameter management.

[0144] Specifically, the parameter storage and retrieval system of the breast pump introduces a user star filtering strategy, which can accurately match the user's own lactation needs priority and solve the problem that the general parameter retrieval and sorting logic cannot accurately identify the user's personal key preferences.

[0145] Users can actively annotate snapshots of core lactation historical parameters that are comfortable for them and have high milk production efficiency by starring them. During the lactation parameter retrieval and matching process, the system will prioritize locking the starred parameter snapshots that meet the search conditions of the current lactation period and mammary gland status. Users do not need to search through a large number of ordinary historical parameter snapshots that meet the basic conditions one by one to locate them, which greatly shortens the path for users to match suitable parameters and improves the recall efficiency of target suitable parameter snapshots.

[0146] The above filtering logic fully respects the user's personalized lactation status classification, placing the highly relevant key parameters actively marked by the user at the highest priority level of the filtering. This better adapts to the user's custom management habits of their personal lactation history, avoiding the situation where highly relevant personal parameters are overwhelmed by a large number of low-relevance ordinary parameter snapshots. This effectively improves the accuracy of parameter retrieval and matching in lactation scenarios and optimizes the user experience of retrieving relevant parameters.

[0147] The user star-marking filtering strategy can also complement the preset filtering strategies such as lactation period and milk production stage to form a hierarchical filtering logic. When there is a snapshot of the star-marked parameter that meets the current conditions, the target parameter is directly hit. This not only retains the flexibility of multi-dimensional filtering, but also enhances the personalized adaptability of parameter matching results through the user's active marking. This allows the entire parameter retrieval and filtering system to have both general scenario rule processing capabilities and personalized lactation demand response capabilities, further enhancing the usability of the entire breast pump parameter storage and retrieval system.

[0148] Parameters are dynamically and adaptively adjusted based on real-time lactation data. After the initial screening and matching of starred parameters are completed, the real-time parameter adaptive adjustment process is initiated: First, the lactation stimulation parameters actively input by the user during the current lactation period are acquired. Simultaneously, dynamic data of mammary gland pressure at the current lactation stage are collected. The negative pressure and stimulation frequency of the milk pump are dynamically adjusted in combination with the collected real-time data. The operating parameters of the breast pump that match the current real-time lactation status are output, adapting to the dynamic changes in mammary gland status and milk flow rate during lactation, avoiding problems such as mammary gland discomfort or low milk flow efficiency caused by fixed parameters.

[0149] The system recommends optimal parameters for the same scenario, adapting to user manual adjustments. Based on a user-starred selection system, a historical lactation event snapshot database can be built. This database extracts historical validation data for similar lactation scenarios, summarizing the optimal parameter range for each scenario. When a user needs to manually adjust the breast pump parameters, the system automatically provides parameter recommendations that align with the user's milk production patterns, effectively reducing the trial-and-error costs and operational complexity of manual adjustments, and further improving the accuracy of parameter adaptation and user experience.

[0150] In some examples, the process of loading the parameter vector of a target event snapshot includes: First, read the firmware version of the current device and the model of the currently connected pump body, and then compare them with the original firmware version and the original pump body model recorded in the target event snapshot. If there are version differences, the original parameter number is mapped to the parameter number corresponding to the current firmware version through a pre-stored version mapping table.

[0151] The above verification and matching steps can effectively solve the problem of incompatibility of old lactation event snapshot parameters caused by firmware updates and hardware parameter adjustments during the product iteration and upgrade process of breast pumps through the pre-verification and matching mechanism of breast pump firmware version and pump hardware model. This avoids failures such as snapshot loading failure and lactation event data parsing errors caused by parameter number mismatch, and greatly improves the compatibility and traceability of historical event data between different version iterations of products. The parameter number can be automatically adapted and converted based on the pre-stored version mapping table. There is no need to manually modify or re-enter historical lactation event snapshot data. This not only preserves the integrity of the original lactation event records, but also greatly reduces the data adaptation cost during after-sales maintenance and equipment upgrades. It can quickly load and parse cross-version lactation event snapshots to meet the functional requirements of breast pump equipment failure tracing and usage event review. This adaptation logic relies on a direct comparison of hardware models and firmware versions, eliminating the need for complex conversion calculations on the original parameter vectors. The adaptation process is simple and efficient, without adding extra computational load to the breast pump device. It ensures the response speed of snapshot loading, adapts to the limited computational resources of embedded breast pump devices, and effectively improves the stability of system operation.

[0152] In some examples, the steps after obtaining the mapped parameters also include: cropping and scaling the snapshot parameters according to the parameter boundaries of the current pump hardware to obtain a parameter vector adapted to the current hardware, which serves as the current operating parameters of the breast pump.

[0153] By cropping and scaling the snapshot parameters to fit the current pump hardware, it can be ensured that the output operating parameters always fit the parameter boundary range of the target hardware, avoiding problems such as breast pump malfunction and component overload damage caused by parameters exceeding the limit, and effectively extending the service life of the breast pump hardware. The adapted parameters are perfectly matched to the actual conditions such as the current pump body's structural characteristics and load capacity, which can keep the breast pump in the optimal working state at all times. This ensures that the negative pressure suction and massage frequency meet the lactation needs, improves milking efficiency and comfort, and avoids useless power output, reducing the overall energy consumption of the product. The above adaptation method is compatible with pump hardware of different specifications and parameters. The same core control algorithm can be adapted to multiple models of breast pump products without the need to redevelop and train parameter models for different hardware. This greatly reduces the product development and mass production debugging costs and effectively shortens the new product launch cycle. In addition, when individual parameters of breast pump hardware deviate due to manufacturing tolerances, this process can quickly correct and adapt the parameters, improve the performance consistency of products in the same batch, and reduce the defect rate.

[0154] In some examples, the structured parameter vector of the event snapshot includes at least: vacuum level parameter, milk suction frequency parameter, and suction duty cycle parameter.

[0155] The above method integrates vacuum parameters, pumping frequency parameters, and suction duty cycle parameters into a structured parameter vector to construct an event snapshot. This accurately captures the actual pumping needs of different lactation stages and individuals, solving the problem that traditional breast pumps with fixed parameters cannot adapt to individual physiological differences in lactation. This significantly improves the comfort and adaptability of using the breast pump. It achieves precise matching based on the physiological characteristics of different lactation stages. In the early stages of lactation, the mammary glands are highly sensitive and the milk secretion is low. The structured parameter vector can record the comfortable stimulation mode parameters of low vacuum, low frequency, and high suction duty cycle. During the peak lactation period, the milk is plentiful and higher emptying efficiency is required. It can automatically switch to match the suction duty cycle parameters of high vacuum, medium to high frequency, and adapted to the lactation rhythm, which not only ensures the effect of stimulating lactation, but also avoids mammary gland damage caused by unsuitable parameters, and reduces the risk of nipple cracking, negative pressure damage to the mammary glands, etc.

[0156] It enables efficient storage and reuse of personalized parameters. For users who regularly use breast pumps, there is no need to manually adjust parameters every time they use the device. The stored structured parameter vectors can be directly matched with the user's past comfortable pumping patterns, significantly reducing the time cost of parameter adjustment and improving the user experience.

[0157] This provides an accurate data foundation for the intelligent parameter adjustment of breast pumps. Based on structured parameter vectors, the breast pump can identify changes in the user's current state during the pumping process in real time and dynamically adjust various parameters. While ensuring efficient milk emptying, it maximizes the simulation of the baby's sucking rhythm during breastfeeding, better stimulates the milk ejection reflex, maintains a stable milk supply, and can also adapt to the needs of users with different tolerance levels, balancing pumping efficiency and user comfort.

[0158] The standardized parameter structure also provides a unified data format for subsequent functional iterations of breast pumps and analysis of user lactation behavior, making it easier for products to optimize parameter logic based on large sample user data, and further improving the adaptability and intelligence of the products.

[0159] In some examples, the event snapshot includes at least one of the following tag data: The data includes the following information: milk pumping side identifier, data collection period label, firmware version number at the time of data collection, pump model identifier at the time of data collection, user star mark, milk volume gain parameter at the time of data collection, and parameter confidence level parameter.

[0160] The above-mentioned data can be structured and labeled across all dimensions of the breast pump's operation process by configuring one or more of the following labels for the event snapshot: milk pump side identifier, collection period label, firmware version number at collection time, pump model identifier at collection time, user star mark, milk volume gain parameter at collection time, and parameter confidence parameter. This provides a clearly labeled sample dataset for subsequent parameter optimization model training. Based on the differences in lactation characteristics of different breast pumping sides (single / double pumping, left / right pumping), the negative pressure and rhythm parameters of the corresponding breast pumping side can be adjusted accordingly based on the breast pumping side label. This solves the problem of poor adaptability of uniform parameters when the lactation sensitivity of the two sides is different, and can take into account the lactation comfort of different breasts, avoiding overstimulation or insufficient lactation on one side. By collecting time period tags to match the needs of breast pump users at different lactation stages, it can automatically match parameters for different stages such as the lactation initiation period, regular lactation period, milk replenishment period, and weaning period, without requiring users to manually adjust repeatedly, greatly reducing the barrier to entry for using breast pumps. By combining the firmware version number and pump model identifier during data collection, it can accurately adapt to breast pump products with different hardware specifications and software versions, avoid output characteristic deviations when migrating parameters across models and versions, ensure the stability and consistency of parameter control on different hardware platforms, and provide accurate comparison data for parameter iteration of older firmware versions, supporting rapid product iteration and optimization. User star marking can retain user-marked combinations of comfortable and efficient breast pumping parameters, transforming subjective user experience into reusable high-quality training samples. This makes the parameter control model more aligned with individual user habits, enabling rapid learning and memorization of personalized parameters. In subsequent use, the optimal parameter combination tailored to the user can be directly retrieved, improving the user experience. By collecting milk volume gain parameters and parameter confidence parameters, the lactation effect of the current parameter combination can be quantitatively evaluated. High-quality parameters with high lactation efficiency and good user experience can be screened out, while parameter combinations with low confidence and poor gain effect can be eliminated. The parameter control logic can be continuously optimized to improve milk pumping efficiency. While ensuring milk volume, the duration of a single pumping session can be shortened. Confidence screening can also reduce the interference of abnormal data collection on the parameter model and improve the robustness of parameter control.

[0161] In summary, this disclosure, through the labeling of multi-dimensional tag data, can not only support the continuous iterative optimization of the breast pump parameter control model, but also achieve personalized and precise parameter adaptation for different users, different hardware, and different lactation stages, ultimately improving breast pumping efficiency, enhancing breast pumping comfort, and reducing user operating costs.

[0162] Reference Figure 3 In some examples, before the step of responding to a breast pump start request and matching the target event snapshot from the snapshot storage collection according to a preset filtering strategy, a side-by-side and time-segmented bucket retrieval step is also included: Obtain the milk pumping side identifier and the time period to which the current milk pumping request belongs, and match event snapshots from the snapshot storage set that have the same milk pumping side identifier and the same time period label to participate in the filtering.

[0163] The above-mentioned side-by-side and time-by-time bucket search steps can accurately limit the search scope based on the differences in lactation patterns of different breasts. This is in line with the physiological characteristics of most users, such as the natural differences in lactation volume and lactation sensitivity between the left and right sides, and the regular fluctuation of human lactation volume with the time of day and night. It greatly reduces the amount of calculation of irrelevant snapshot data, improves the matching efficiency of target event snapshots, avoids interference from historical lactation data from different sides and different time periods with parameter matching results, and makes the matched working parameters more in line with the actual lactation status of the current breast at the current time. Bucket-based retrieval categorizes and stores scattered historical event snapshots by the pumping side and time period, which reduces the traversal cost of a single retrieval, improves the response speed of breast pump parameter matching, and better adapts to the historical data accumulated by users over a long period of use. This allows the accuracy of parameter matching to continuously improve with the number of uses, reducing the operational cost of users manually adjusting parameters and avoiding problems such as insufficient milk production and nipple pain caused by parameter mismatch, thus balancing pumping efficiency and pumping comfort.

[0164] Furthermore, the content of the side-by-side independent snapshots and differential matching of the left and right breasts includes: In scenarios involving dual-sided breast pumps or partial single-sided use, the milk production characteristics of the mother's left and right breasts may differ. In this embodiment, the snapshot record carries a left / right identifier (Side).

[0165] When a milk let-down event occurs on the left side, a snapshot P_L = {Side=L, N_set=Level 4, f_pwm=55, D_inhale=50%} is written to storage; when it occurs on the right side, P_R = {Side=R, N_set=Level 6, f_pwm=65, D_inhale=58%} is written. The next time the user selects the left side, the controller only loads P*_L; the right side loads P*_R. When both sides are used simultaneously, the left and right pumps load their respective snapshots and perform a smooth transition independently.

[0166] Furthermore, the snapshots taken at different times, adapting to the differences in lactation between morning and evening, include the following content: The milk production and let-down characteristics of breastfeeding mothers vary throughout the day (e.g., milk production is usually higher in the early morning). In this embodiment, the snapshot record additionally carries a time band (e.g., "0-8 am", "8 am-4 pm", "4 pm-12 am"). Upon the next power-on, the main controller determines the current time period based on the system clock and prioritizes loading the best snapshot P* for that time period. If there is no snapshot for the current time period, it reverts to the best snapshot for the entire time period or the default program.

[0167] In some examples, when no event snapshot matches the current breastfeeding side identifier, a separate search for each side is not triggered, and a global snapshot is used for filtering. If no event snapshot matches the time period label of the current time, time period matching is not triggered, and all snapshots of the corresponding breast pumping side are used for filtering.

[0168] The adaptive retrieval strategy based on the above branch judgment can flexibly adjust the retrieval scope according to the different data coverage of the actual use scenario of the breast pump. It retains the advantages of accurate matching by pumping side and time period, while avoiding the failure of the recommendation model due to the lack of corresponding matching snapshots, ensuring that the breast pumping parameter recommendation process can always run stably. The accuracy of personalized adaptation for breast pumping habits has been further improved. For users whose milk secretion patterns differ between different breasts, even if there is insufficient milk secretion records on one side, parameters that match the user's overall milk secretion characteristics can be given based on global milk secretion data without breaking the logic of differentiated adaptation for each side. It will not force the uniformity of milk secretion parameters on both sides due to the lack of data on one side, thus balancing personalized adaptation and data continuity. The system is more robust in adapting to the differences in lactation patterns at different pumping times. For users who have just started pumping regularly and have not yet accumulated full-time usage data, it will not fail to provide matching parameters because there is no historical data for the target time period. Based on the full-time historical data of the corresponding pumping side, it can still output parameter schemes that conform to the lactation characteristics of that side, which greatly improves the initial user experience and avoids the discomfort of repeated parameter adjustments in the novice stage. The layered and progressive retrieval and matching logic controls the risk of overfitting in small data scenarios and avoids the waste of computing power caused by full-scale retrieval: when a corresponding matching snapshot exists, the precise matching path is followed to ensure the accuracy of parameter recommendations; when a match is missing, the fallback adaptation path is followed to ensure the availability of the process. This allows the breast pump's automatic parameter adjustment function to stably output the suction power, stimulation frequency, and pumping rhythm combination that meets the user's needs in various usage scenarios, effectively improving lactation efficiency, reducing the risk of nipple damage caused by improper negative pressure over a long period of time, and better adapting to the lactation patterns of different scenarios and different sides of the user, thus improving the comfort of breast pumping.

[0169] Reference Figure 3 The attached diagram illustrates the side-by-side and time-based bucket retrieval logic block diagram in the parameter control method of a breast pump. The process is broken down step-by-step as follows: Session starts, reads user ID. When a user powers on the device or initiates a breast pumping session via the app, the main controller reads the currently logged-in user ID from the memory, isolates snapshot data from different users, and prevents parameter corruption when multiple users share the device.

[0170] Branch determination: This breast pumping session is left side L / right side R. When storing snapshots, each record carries a Side (left / right breast identifier). The controller first distinguishes the pumping side and stores it in a separate data bucket. Select left breast pumping → Enter the Side=L dedicated snapshot bucket to retrieve only all milk let-down snapshots generated by the left breast; Select right-side breast pumping → Enter the Side=R dedicated snapshot bucket to retrieve only all milk let-down snapshots generated by the right breast.

[0171] Corresponding to Patent Example 4: The left and right breasts have different lactation characteristics, and each side takes an independent snapshot. When both sides are used at the same time, the left and right pumps are retrieved independently.

[0172] Secondary filtering within the bucket: Matching the current time period (morning / noon / evening) After entering the single-sided snapshot bucket, the current time period (0-8 am, 8-16 pm, 16-24 pm) is determined based on the device system clock, and only snapshot subsets with the same time period label are filtered; Corresponding to patent embodiment 5: There are large differences in milk production and milk let-down response at different times during lactation, with vigorous milk production in the early morning and weaker milk production in the evening. Time-segmented filtering further improves the matching accuracy.

[0173] Weighted sorting of snapshots within the bucket, output the best snapshot P.

[0174] For multiple snapshots from the same user, same side, and same time period after filtering, calculate the weighted score W=λ1. Score+λ2 normalize(milk_gain)+λ3 Sort user_rating from high to low, and select the top 1 as the target snapshot P. It supports three selection strategies: latest priority, weighted centroid average, and priority of user-defined apps manually starred.

[0175] Unified post-processing: Hardware boundary trimming + smooth parameter transition The left and right sides retrieve their respective optimal snapshots (PL). PR Then, it enters a unified processing flow: Hardware boundary trimming: Limits the negative pressure and frequency within the snapshot to the maximum negative pressure Nmax and frequency range of the local hardware to prevent damage and breast pain caused by exceeding the hardware limit. Smooth transition: The device does not jump directly to the snapshot high negative pressure upon startup. Instead, it uses a linear / S-curve to gradually approach the target parameters in segments, ensuring comfortable adaptation for the breast. When pumping milk from both sides simultaneously, the left and right pumps are loaded with their respective P. Independently execute smooth adjustment.

[0176] In some examples, cross-version compatibility mapping steps are also included: When loading the parameters of the event snapshot, read the firmware version of the current device and the model of the currently connected pump body, and compare them with the original firmware version and the original pump body model recorded in the event snapshot. If there are version differences, the original parameter number will be mapped to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

[0177] The above steps can achieve seamless compatibility of breast pump parameter event snapshots between different firmware versions and different pump hardware models through cross-version compatibility mapping steps. This avoids the problem that old parameter snapshots cannot be loaded and used due to version iteration and hardware updates. Users do not need to manually set parameters again, which greatly improves the user experience after changing devices and upgrading firmware. By using a pre-stored firmware version mapping table to complete the mapping and matching of parameter numbers, it is possible to quickly adapt to version differences. This eliminates the need to store redundant parameter information for all versions, ensures the accuracy of parameter mapping, reduces the memory footprint of parameter storage, and improves the processing efficiency of parameter adaptation. The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump hardware to obtain a parameter vector adapted to the current hardware. This not only preserves the user's original parameter settings such as lactation mode, massage intensity, and suction frequency, but also avoids problems such as abnormal operation and component damage caused by parameters exceeding the working range of the current hardware. This ensures that the breast pump can operate stably under different hardware configurations and is in line with the user's usage preferences. This compatibility handling method does not require any changes to the hardware structure. Cross-version compatibility can be achieved solely through parameter mapping adaptation at the software level. This reduces the development and adaptation costs of hardware iterations and provides a more flexible parameter compatibility solution for breast pump function upgrades and hardware updates, thus extending the product ecosystem lifecycle.

[0178] Specifically, the snapshot compatibility mapping for firmware upgrades and pump body replacements includes: When the device upgrades its firmware via OTA, the lactation program numbers may be rearranged (e.g., the old version "Program 3" corresponds to the new version "Program 5"). A firmware version mapping table is maintained in storage 50. When loading a snapshot, the controller automatically maps the old program number to the new version number; if no mapping relationship exists, it reverts to the closest default curve family in the new version and prompts the user for confirmation.

[0179] When a user replaces the pump body with a different model (e.g., from a standard model to a silent model), the maximum negative pressure N_max_hw of the new pump body may decrease. During automatic matching, the main controller performs hardware boundary trimming: N_set = min(snapshot N_set, new N_max_hw), scaling the frequency and duty cycle proportionally or keeping them unchanged to ensure they do not exceed the capabilities of the new hardware.

[0180] In some examples, parameter boundary trimming includes: if the vacuum parameter in the event snapshot is higher than the current pump body hardware's preset negative pressure safety limit, then the vacuum parameter is automatically adjusted to the negative pressure safety limit, while preserving the proportional relationship of the other parameters.

[0181] By using the above parameter boundary trimming rules, it is possible to ensure that the overall negative pressure output characteristics of the current lactation mode of the breast pump match the user's customized lactation rhythm, while avoiding the hardware from working in an overload state that exceeds the threshold for a long time. This not only prevents excessive negative pressure from causing unnecessary stimulation and damage to the user's mammary glands, but also avoids the risk of aging acceleration and sealing failure of the pump motor and air circuit structure due to overpressure operation. At the same time, the proportional relationship of other parameters is retained, without the need to recalculate the matching relationship of the entire set of lactation parameters. This will not disrupt the lactation stimulation-pumping cycle rhythm that the user has already adapted to, ensuring the continuity and comfort of the pumping experience. It also reduces the computing power required for parameter control, allowing even low-configuration main control chips to achieve rapid response and adjustment.

[0182] The aforementioned negative pressure limit adjustment logic is a passively triggered adjustment, which only takes effect when the parameters manually set by the user or generated by mode iteration exceed the safety threshold. It will not interfere with the lactation process within the normal parameter range. It retains the flexibility for users to customize the adjustment parameters according to their own lactation situation and tolerance, while also improving the safety of product use through an invisible safety backup mechanism. It adapts to the usage needs of users with different physical conditions and different lactation stages. Whether it is insufficient milk supply requiring increased negative pressure stimulation or high sensitivity requiring control of the negative pressure range, it can provide safety assurance while making personalized adjustments.

[0183] Some examples also include the milk array window linkage control steps: When a milk ejection event is detected to enter the milk ejection duration window, a short-term enhancement adjustment is triggered, and the current operating parameters are adjusted to the enhancement parameters. When generating an event snapshot, the running parameters are extracted from the moment after the milk flow occurs and before the enhanced adjustment takes effect. The enhanced parameters after the short-term enhanced adjustment are not recorded.

[0184] By setting up a milk let-down window linkage control step, parameters can be adjusted specifically during the milk let-down stage to enhance milk expression efficiency. This not only matches the physiological rhythm of lactation and shortens the milk expression time, but also avoids excessive breast stimulation caused by maintaining high power parameters throughout the process, thus improving user comfort. By generating an event snapshot by extracting only the operating parameters after the milk let-down occurs and just before the enhanced adjustment takes effect, without recording the enhanced parameters after the short-term enhanced adjustment, the normal baseline parameters of lactation during milk expression can be preserved. This avoids interference from single short-term special adjustment parameters with the subsequent parameter self-matching logic, ensuring that the device can generate an adaptation scheme based on the user's normal lactation characteristics when it makes subsequent automatic adjustments. This improves the adaptability and accuracy of the parameter control model, while also reducing the storage of invalid parameters and optimizing the computing power consumption of the device. The above parameter control logic can automatically identify the lactation stage and complete dynamic adaptation, eliminating the need for users to manually adjust parameters repeatedly. This lowers the barrier to entry for using the breast pump, making it more user-friendly for novice users, while also balancing pumping efficiency, user comfort, and device control stability.

[0185] Specifically, real-time mode enhancement and snapshot collaboration within the milk array window can include the following: After detecting a milk ejection event, in addition to performing snapshot capture, the main controller can also make short-term enhancement adjustments to the current mode within the milk ejection duration window to make full use of the milk ejection peak: It automatically increases the inhalation duty cycle D_inhale by 2-5 percentage points for a short period of time, extending the effective inhalation time per cycle; Alternatively, insert 2 to 3 "deep suction pulses," increasing the negative pressure by 10% within a single cycle, before returning to the normal value.

[0186] Key linkage principle: The parameter P written to the snapshot should be the regular parameter "moment before the deep suction pulse", rather than the enhanced temporary parameter. This way, the next automatic matching will be the user's long-term comfortable and acceptable operating point, avoiding the incorrect solidification of excessively high negative pressure into the normal state.

[0187] In some examples, a snapshot incremental update step is also included: When a user manually adjusts the running parameters in the current lactation session, a confirmation prompt is output asking whether to update the snapshot set, and the user's confirmation response is received. If the user confirms the update, the adjusted parameters will be written as a new event snapshot based on the current capacity of the snapshot storage set. When the storage capacity exceeds the limit, the event snapshot generated earliest will be replaced. If the user confirms that they will not update, the adjusted parameters will be valid for the current lactation session and will not be written to the snapshot storage set.

[0188] By introducing a snapshot-based progressive update process, it is possible to accurately adapt to the personalized parameter adjustment needs of different users in the same lactation session, respecting the user's autonomy while continuously optimizing the data source for intelligent parameter recommendations.

[0189] When users choose not to update the snapshot set, the temporary adjustment needs of a single lactation session can be met, and occasional parameter adjustments will not interfere with the historical snapshot data that has been accumulated and conforms to the user's long-term usage habits, thus ensuring the stability of subsequent parameter recommendations. When a user confirms an update to the snapshot set, the snapshot set can be dynamically maintained based on storage capacity. By eliminating the oldest snapshot generated at the earliest time, the snapshot set always ensures that it stores the latest parameter data that is in line with the user's current lactation habits, avoiding excessive storage capacity occupation. At the same time, it makes the initial parameter recommendations when starting subsequent lactation sessions more in line with the user's current physical condition and usage needs, continuously improving the user experience. The aforementioned update mechanism balances the need for storage resource consumption with the accuracy of intelligent recommendations. It avoids wasting storage space due to unlimited storage snapshots, while continuously optimizing parameter adaptation capabilities through dynamic updates. This allows the breast pump's parameter control to retain flexible adjustment space and gradually achieve more precise personalized adaptation.

[0190] In some examples, when the user confirms the update, it is determined whether the difference between the currently manually adjusted parameters and the target parameters in the existing snapshot exceeds a preset difference threshold. If it does not exceed the preset difference threshold, the weight parameters of the corresponding existing snapshot are updated directly without adding a new snapshot entry.

[0191] By determining whether the difference between the manually adjusted parameters and the existing snapshot target parameters exceeds a preset difference threshold, the existing snapshot weight parameters are updated directly without adding new snapshot entries only if the difference does not exceed the threshold.

[0192] The above processing method can effectively reduce the redundant generation of parameter snapshot entries, avoid the accumulation of a large amount of duplicate snapshot data with very small parameter differences in the storage device, reduce the storage resource occupancy rate, and improve the operating space utilization efficiency of the breast pump parameter storage system. The streamlined snapshot library can shorten the time spent on subsequent parameter retrieval and matching. When users call the historical lactation pattern again, the system can locate and match the corresponding parameter snapshot more quickly, improve the response speed of pattern retrieval, and optimize the user experience.

[0193] The above processing method preserves the user's need for manual parameter adjustment and updates, without requiring the user to perform additional operations such as deleting duplicate snapshots and organizing the repository. While meeting the user's parameter adjustment habits, it reduces the user's operational burden and ensures that the parameter snapshot library always maintains a clear and reasonable structure, providing stable support for the long-term storage and retrieval of personalized lactation modes of breast pumps.

[0194] Controlling the number of snapshot entries can also reduce the computational workload of subsequent parameter matching iterations, reduce the operation load of the built-in control chip of the breast pump, help reduce the device power consumption, extend the battery life after charging the breast pump, and improve the overall stability of the device operation.

[0195] Specifically, the snapshot update mechanism after the user manually adjusts parameters can include: After the automatic matching is completed and enters the steady state of snapshot parameters, if the user manually makes a large adjustment to the parameters during this session (for example, lowers the negative pressure from the 5th gear of the snapshot to the 3rd gear), it indicates that there is a deviation between the user's current state preference and the historical snapshot. At this time, the main control triggers the following linkages: ① Pop up a prompt "It is detected that you have adjusted the parameters. Do you want to update the current settings to the new preference?"; ② If the user confirms, write the adjusted parameter P_new into the storage as the new snapshot (or replace the oldest one); ③ If the user refuses or does not operate, then this parameter adjustment is only valid for the current session and does not affect the historical snapshot.

[0196] The above mechanism enables the snapshot set to be gradually updated as the user's lactation state changes, always reflecting the latest preference.

[0197] In some examples, it also includes the snapshot processing steps in the special safety mode: When entering the special safety mode, perform a safety audit on all event snapshot parameters to be loaded in the snapshot storage set, forcefully clip the parameters that exceed the preset parameter boundaries of the special safety mode to within the boundary values, add an audit flag to the clipped event snapshots, and load the clipped parameters without discarding the original snapshot data.

[0198] The above processing steps of performing safety audits, boundary clipping, and adding audit flags to all event snapshot parameters to be loaded in the special safety mode can achieve the following beneficial effects: Ensure the safe operation of the breast pump and avoid safety risks: For special scenarios such as sudden abnormal working conditions like abnormal battery voltage, excessive motor load, and excessive pumping pressure, by forcefully clipping the out-of-bounds parameters to within the safe boundaries, it not only avoids problems such as the breast pump motor overload, pressure out-of-control, and circuit overcurrent caused by out-of-range parameters, but also does not need to directly trigger shutdown and locking, enabling the breast pump to continue operating within the safe power and safe pressure ranges, which not only ensures the user's safety in use but also avoids sudden interruption of milk pumping and affecting the user experience.

[0199] The original data is fully preserved to support product optimization: This processing method does not discard the original abnormal snapshot data, but only trims and adjusts the operating parameters. At the same time, a special audit mark is added to distinguish the modified parameters from the original parameters. This will not affect the normal operation and control of the current breast pump, and can also fully preserve the original user data and operating parameters under abnormal scenarios. The original snapshot data can be read through cloud synchronization or local access in the future to analyze the triggering cause of abnormal operating conditions and user habits, providing complete and true original data support for the optimization of the breast pump's parameter control logic and product iteration and upgrade.

[0200] Enhancing adaptability to abnormal scenarios while balancing safety and continuity of use: Unlike traditional methods that directly refuse to load abnormal parameters and trigger device shutdown, this method can still load the trimmed compliant parameters to maintain device operation while completing security audit and control. This satisfies the risk control requirements of special security modes without interrupting the user's ongoing breast pumping process, avoiding issues such as breast milk waste and decreased user experience caused by sudden shutdowns. It significantly improves the user experience and usability of the breast pump under abnormal operating conditions.

[0201] Facilitates subsequent source tracing and investigation, and improves after-sales maintenance efficiency: The added audit flags can quickly locate abnormal snapshots where parameters have exceeded limits. During after-sales fault investigation and user problem tracing, it can quickly identify the occurrence scenario and original parameters of abnormal operating conditions, helping maintenance personnel to quickly locate the cause of the problem without having to search the entire data one by one, effectively reducing investigation costs and improving after-sales processing efficiency.

[0202] In some examples, the special safety mode is a nipple hypersensitivity mode or a premature infant breastfeeding mode, and the upper limit of negative pressure of the preset parameter boundary of the special safety mode is lower than the upper limit of negative pressure of the normal mode.

[0203] By setting the preset negative pressure limit of the special safety mode to be lower than that of the normal mode, targeted safety protection can be provided for the special needs of users with highly sensitive nipples and for breastfeeding scenarios of premature infants.

[0204] For breastfeeding mothers with sensitive, cracked, or swollen nipples, a lower upper limit of negative pressure can prevent excessive negative pressure during breastfeeding from causing additional pulling and damage to the already sensitive nipple and areola tissue. This can alleviate the stinging and discomfort during breastfeeding, allow space for the nipple wound to heal, reduce the risk of nipple ulceration and cracking worsening, and improve the tolerance and comfort of the breastfeeding process. For premature infants breastfeeding, a lower upper limit of negative pressure can adapt to the fragile sucking ability of premature infants and the state of the breasts in the early stages of lactation. It simulates the gentler, more natural sucking rhythm of premature infants, which can avoid milk waste caused by the milk flowing out too quickly during the feeding process. It can also steadily stimulate prolactin secretion, gradually establish a stable milk ejection reflex, and ensure that the collected milk can meet the slow feeding needs of premature infants. It can also help mothers gently clear the milk ducts in the early postpartum period of premature birth, and avoid damage to breast tissue caused by excessive negative pressure.

[0205] In addition, fixed parameter boundary limits can prevent users from accidentally increasing the negative pressure beyond the safe range. The system can automatically ensure the safety of the breast pumping process without requiring users to repeatedly adjust the negative pressure parameters themselves. This reduces the operational threshold and health risks for special groups of people using breast pumps, and improves the product's adaptability to different scenarios and the safety and user experience of using it.

[0206] Furthermore, when the device (breast pump) is in the "premature / highly sensitive" safety mode, its hardware negative pressure limit N_max_hw will be dynamically reduced by the system, for example, from the usual -250mmHg to a more conservative -150mmHg. The core of this mechanism is that even if the user's custom-set negative pressure value N_set has exceeded this safety limit in their historical usage snapshots, the system will forcibly trim the recommended parameters to the safety limit value for this mode during automatic parameter matching and recommendation, and record the corresponding audit operation flag in the background. Simultaneously, to ensure users are fully informed and enhance their sense of security, the human-computer interaction module and / or the connected mobile application APP will simultaneously display clear prompts on the interface, such as "Negative pressure has been reduced due to safety mode." These settings ensure that, in usage scenarios targeting mothers of premature infants or users with highly sensitive breast tissue, the intelligent recommendation mechanism based on historical snapshots can provide a personalized experience while strictly adhering to and prioritizing the implementation of preset safety limits. This effectively prevents the snapshot mechanism from unintentionally bypassing safety protection, comprehensively ensuring the breast pumping safety and comfort experience for special populations.

[0207] In some examples, a cold start process is triggered when no event snapshot corresponding to the current search criteria exists in the snapshot storage collection. The cold start process includes: Obtain the group prior parameter vector of users with the same breast pump model and nipple size from the cloud, and use the group prior parameter vector as the initial recommendation parameters. After each new individual event snapshot is generated, the fusion weight of the group prior parameters is reduced; When the number of individual event snapshots reaches the preset cold start threshold, the fusion weight of the group prior parameters is adjusted to zero, and the filtering results of individual snapshots are used.

[0208] The above-described cold start process effectively addresses the pain point of inaccurate parameter recommendations and the need for users to repeatedly manually adjust the settings when using a breast pump, due to the lack of individual historical usage data.

[0209] Initial recommendations are made based on prior parameters of the same model of breast pump and the same nipple size group. The initial recommended parameters are in line with the general usage habits of this group. Compared with random initialization parameters or fixed default parameters, the initial comfort and lactation efficiency are better, which greatly reduces the frequency and operational burden of new users adjusting parameters in the early stage and improves the experience of first-time users.

[0210] During the cold start process, the fusion weight of the group's prior parameters is dynamically adjusted. As individual user data accumulates, the influence of group parameters on the recommendation results is gradually reduced. This approach can ensure the stability of parameter recommendations by leveraging group experience when data is limited, and can also continuously shift towards personalized user needs as individual data increases. This gradually improves the adaptability of parameter recommendations, achieving a smooth transition from general adaptation to personalized matching and avoiding significant fluctuations in parameter recommendations that could negatively impact the user experience.

[0211] Once the number of individual event snapshots reaches a preset threshold, the system switches entirely to the filtered results of individual data. At this point, the individual data is sufficient to support accurate personalized parameter recommendations, completely breaking free from the limitations of general parameters for the group. This allows the parameter recommendations to perfectly match the user's own lactation patterns, pain tolerance, and usage habits. In the long run, this can stably output parameter control results with higher matching accuracy, improve breast pumping efficiency, reduce discomfort, and at the same time reduce the frequency of manual adjustments by the user, thus optimizing the usage process.

[0212] The group prior parameters stored in the cloud are updated based on the usage data of a large number of users with similar characteristics, which can continuously reflect the general optimal needs of this user group. It does not require a large number of general parameters to be stored on a single device, which reduces the storage burden on the local end of the breast pump and ensures the timeliness and accuracy of the initial recommended parameters. At the same time, the cold start process runs automatically based on the event snapshot mechanism, which does not require users to actively complete personalized configuration operations, making it more accessible and user-friendly for non-technical users.

[0213] In some examples, the group prior parameter vector is the cluster centroid parameter obtained by clustering valid event snapshots of users of the same model and size in the cloud.

[0214] The aforementioned group prior parameter vector, obtained by clustering effective event snapshots of users of the same model and size, serves as the initial benchmark for breast pump parameter control. This approach aligns with the general lactation patterns and usage preferences of the target user group, avoiding the lengthy trial-and-error process of adapting a brand-new device to individual users from scratch. It significantly shortens the waiting time for users to obtain suitable parameters and reduces the probability of breastfeeding pain and low lactation efficiency caused by parameter mismatch during the initial use of new users.

[0215] The centroid parameters obtained by clustering large-scale group data integrate the effective usage experience of a massive number of users in the same scenario. This avoids the limitations of parameter adaptation caused by the bias of individual user habits, and provides a more scientific and reasonable starting optimization direction for the dynamic adjustment of individual parameters. It improves the speed of iterative convergence of individual parameters, allowing the breast pump to adapt to the personalized characteristics of different users such as breast structure, milk production, and pain tolerance more quickly. This not only ensures the efficiency of milk emptying, but also improves the comfort of users during use, reduces the computing resource consumption of the breast pump parameter control module, and improves the response speed of parameter adjustment. At the same time, it provides a continuously updated data foundation for the subsequent optimization of parameter benchmarks of similar products in the cloud, and can continuously iterate and improve the user experience of all users.

[0216] Furthermore, this disclosure also features a cold start fusion mechanism that integrates cloud-based group priors with individual snapshots.

[0217] When a new user uses the system for the first time, there is no individual snapshot of that user. At this time, the communication module reports the user's shield size and pump model to the cloud. The cloud filters out the cluster centroids of users with the same model and size from the accumulated anonymized snapshot data and sends them to the device master controller as the "group prior P_prior". The master controller uses P_prior as the initial P* and performs a smooth transition matching. After the user has accumulated 2-3 individual snapshots, P* will gradually converge to be dominated by individual data, and the weight of the group prior will decay to zero. If the device does not have a communication module, this function will not be available, and the master controller will revert to the factory default program.

[0218] Furthermore, given the industry context where traditional breast pumps generally rely on fixed preset modes or manual user adjustments, resulting in low parameter matching and insufficient lactation efficiency, the breast pump parameter control method proposed in this disclosure achieves personalized and precise adaptation of lactation modes through innovative physiological event anchoring and intelligent snapshot mechanisms. The above control method can achieve the following beneficial effects: 1. When a milk initiation event or milk let-down event is detected, this disclosure can automatically capture and structure and persist the lactation mode-related parameters of the breast pump at that time, forming an "event-mode snapshot" with physiological event anchor points as time markers and bound to the user identity.

[0219] The above solution differs from existing technologies that only support full playback or real-time adjustment. It can accurately record the user's optimal parameter combination during peak lactation, providing a scientific basis for subsequent use.

[0220] 2. When the user uses the breast pump again, the corresponding control system can automatically retrieve matching snapshot data and automatically set or smoothly approximate the lactation mode parameters within the hardware capability boundaries to achieve "hot start" across sessions.

[0221] This feature fundamentally solves the pain point that users have to re-explore the optimal parameters every time they turn on their devices, greatly improving efficiency and comfort.

[0222] 3. The system adopts a multi-snapshot selection strategy, including the latest priority (time-series filtering strategy, prioritizing the latest through time-series filtering), weighted centroid (weighted centroid filtering strategy), and user star (user star filtering strategy), etc. Combined with side / time-segmented bucket retrieval technology, it can automatically match the most suitable lactation mode according to the user's usage habits and physiological cycle.

[0223] As usage increases, the system will continuously optimize matching accuracy, gradually forming a personalized lactation profile.

[0224] 4. To ensure comfort and safety during startup, the corresponding control system can adopt a smooth transition mechanism, using piecewise linear or S-curve approximation algorithms to smoothly transition the breast pump parameters from the initial state to the optimal parameters recorded in the snapshot.

[0225] The above settings effectively avoid discomfort caused by sudden parameter changes and improve the user experience.

[0226] 5. Considering the hardware differences between different breast pump models, the system has a built-in hardware capability trimming and firmware version mapping function, which can automatically identify the device model and firmware version, and make adaptive adjustments to the snapshot data to ensure the compatibility and reuse of snapshots across pump bodies and firmware versions.

[0227] These features greatly enhance the product's versatility and user engagement.

[0228] 6. When the user manually adjusts the parameters, the system will automatically prompt the user whether to update the snapshot data to ensure that the snapshot set can gradually evolve as the user's lactation stage changes.

[0229] The above settings enable the system to continuously adapt to the user's physiological changes and always provide the best service.

[0230] 7. The system prioritizes user experience and data security, providing comprehensive snapshot lifecycle management functions, including user confirmation, one-click undo, and privacy deletion options.

[0231] Users have complete control over their snapshot data, and can view, modify, or delete it at any time, ensuring data security and privacy protection.

[0232] On the other hand, this disclosure also provides a breast pump control system for performing the above-described method. The breast pump control system includes a physiological event detection module, a snapshot generation and storage module, a snapshot filtering and matching module, and a parameter loading module. The physiological event detection module is used to continuously collect the negative pressure value of the breast pump through a pressure sensor, calculate the change in negative pressure per unit time, and determine the occurrence of a milk initiation event or milk let-down event when the change in negative pressure exceeds a preset threshold, or receive a milk let-down marker signal manually triggered by the user. The snapshot generation and storage module is electrically connected to the physiological event detection module. It is used to capture the current operating parameters at the moment a physiological event occurs, generate an event snapshot containing structured parameter vectors and label data, and persistently store the event snapshot in the snapshot storage collection. The snapshot filtering and matching module is electrically connected to the snapshot generation and storage module. Upon receiving a startup request, it matches the target event snapshot from the snapshot storage set according to a preset filtering strategy. When the number of event snapshots that meet the search criteria in the snapshot storage set is less than a preset threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots that meet the search criteria in the snapshot storage set is greater than or equal to a preset threshold, a weighted centroid filtering strategy is used to construct the target event snapshot. The parameter loading module is used to load the parameters of the target event snapshot as the current running parameters. During loading, it first reads the firmware version of the current device and the model of the currently connected pump body hardware, compares the original firmware version and the original pump body model recorded in the event snapshot, and if there is a version difference, it maps the original parameter number to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. Then, it trims and scales the snapshot parameters according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

[0233] In the above structure, the physiological event detection module combines automatic detection and manual triggering for milk ejection reflex recognition, adapting to different user habits and recognition needs in different scenarios: it collects negative pressure data through pressure sensors and determines milk ejection reflexes based on changes per unit time, automatically capturing milk ejection or milk initiation events without user manual operation, reducing the user's operational burden during breast pumping. At the same time, the algorithm logic is simple, has low requirements for device computing power, and does not increase the hardware cost of the breast pump. The design that supports users to manually trigger milk ejection reflex marker signals can deal with special cases of missed or false detections by automatic detection. For example, when some users' milk ejection reflexes arrive, the negative pressure change does not meet the preset threshold, or external interference during breast pumping causes abnormal fluctuations in negative pressure, leading to false detections. Manual marking can correct the deviation of automatic detection, greatly improving the accuracy and flexibility of milk ejection or milk initiation event recognition, and providing a reliable triggering basis for generating accurate event snapshots in the future.

[0234] The snapshot generation and storage module works in conjunction with the physiological event detection module to capture current operating parameters the moment physiological events such as let-down occur, ensuring the timeliness and relevance of the parameters. By generating standardized event snapshots containing structured parameter vectors and label data, the parameter storage format is unified, which facilitates subsequent modules to quickly read and parse parameters and also makes it easier to classify and manage breast pumping parameters for different users and in different scenarios. The design of persistently storing event snapshots in a snapshot storage collection can continuously accumulate individual user breast pumping data, record the user's breast pumping preferences and physiological characteristics at different stages, and allow the breast pump parameters to be continuously optimized to meet individual user needs. This avoids the tedious process of resetting parameters every time the device is started and provides a sufficient data foundation for subsequent personalized parameter matching.

[0235] The snapshot filtering and matching module dynamically selects a matching strategy based on the number of snapshots that meet the criteria, balancing matching efficiency with small datasets and matching accuracy with large datasets. When the number of snapshots meeting the search criteria is less than a preset threshold, a time-series filtering strategy is used to select the latest snapshot as the target snapshot. By default, the more recent the breastfeeding parameters, the more suitable they are for the current user's needs. This simplifies the calculation process with small datasets, quickly obtains matching results, and adapts to the user's usage patterns where needs do not change significantly in a short period of time. When the number of snapshots reaches the threshold, a weighted centroid filtering strategy is used to construct the target snapshot. This integrates parameter information from multiple historical valid snapshots, comprehensively considers the reference value of different historical snapshots, avoids the random bias of a single snapshot, and obtains personalized parameters that better suit the user's current needs. Compared to a fixed filtering strategy, the dynamically adapted strategy selection ensures both computational efficiency and improves the matching accuracy of the target snapshot, balancing device computing power consumption and parameter matching effect.

[0236] The parameter loading module incorporates a firmware version and hardware model adaptation process when loading target snapshot parameters, resolving the incompatibility issue of historical parameters across different device versions and hardware specifications. By mapping parameter numbers through a firmware version mapping table, it can adapt to the needs of breast pump firmware iteration updates. Users can still use previously stored historical parameters normally after upgrading the firmware, without issues such as parameter invalidation or the need for resetting, thus improving backward compatibility and extending the lifespan of historical data. Furthermore, by cropping and scaling snapshot parameters according to the parameter boundaries of the current pump hardware, it can adapt to the hardware capability differences of different pump models. Even if the pump hardware is replaced, historical parameters can be adjusted to fit the current hardware's safety and performance range, ensuring that the parameters function normally. At the same time, it retains the adjustment patterns of historical parameters, eliminating the need for users to reset personalized parameters, significantly improving the hardware compatibility of the breast pump and the user experience.

[0237] The aforementioned breast pump control system features a multi-source milk ejection detection mechanism, balancing the convenience of automated judgment with the adaptability to users' personalized needs.

[0238] The breast pump control system supports collecting negative pressure data through pressure sensors and automatically identifying milk initiation or milk let-down events based on the change in negative pressure per unit time. It can capture physiological events without manual operation by the user, meeting the needs of most users for seamless use. The system retains the option for users to manually trigger milk let-down markers. For users with unique milk secretion patterns who are prone to misjudgment or missed detection by automatic detection, manual marking can ensure the accuracy of event recording, covering usage scenarios for users with different lactation characteristics.

[0239] The tiered event snapshot filtering strategy ensures that the target parameters are adapted to the user's lactation habits, even with limited storage resources.

[0240] For scenarios where the number of snapshots is insufficient in the early stages of user use, the time-series filtering strategy directly selects the most recently generated qualified snapshot to match the user's current lactation status. This avoids the embarrassment of having no available parameters and also fits the user's recent usage habits, ensuring the rationality of parameter loading. Once a certain number of snapshots have been accumulated, a weighted centroid filtering strategy is used to construct a target snapshot. This strategy integrates the parameter features of multiple historical milk-inducing events or milk let-down events, and extracts personalized parameters that best match the user's long-term lactation patterns through weighted calculations. Compared to single snapshot parameters, this approach better suits the user's actual lactation needs, effectively improving pumping comfort and efficiency.

[0241] The parameter loading mechanism with version hardware adaptation solves the parameter compatibility problem between different firmware iterations and pump body hardware of different specifications.

[0242] Before loading snapshot parameters, the breast pump control system first compares the current device with the firmware version and pump hardware model recorded in the snapshot. If there are differences, it first completes the mapping conversion of parameter numbers through a pre-stored version mapping table to avoid parameter reading errors and program crashes caused by changes in firmware parameter rules. By combining the parameter boundaries of the current pump body hardware, the snapshot parameters are cropped and scaled to adapt the historical parameters to the reasonable working range of the current hardware. This not only preserves the user's historical parameter adjustment patterns, but also avoids problems such as insufficient or excessive suction caused by parameter mismatch. This improves the cross-version compatibility and usage stability of the device, while reducing the frequency of users readjusting parameters.

[0243] The structured and persistent event snapshot storage method provides reliable support for the accumulation of personalized user parameters. The event snapshots generated by the breast pump control system contain structured parameter vectors and tag data, which facilitates quick reading, calculation, and processing in subsequent filtering and matching processes. The snapshot collection, persistently stored locally, will not lose user usage data due to device restarts or program updates. It can accumulate user lactation habit data over a long period of time, making the target parameters generated by the system increasingly closer to the user's actual needs. This achieves personalized self-learning of breast pump parameters, providing users with a more comfortable and efficient breast pumping experience compared to fixed multi-level modes.

[0244] Furthermore, the breast pump control system disclosed herein also includes the following components: The Micro Control Unit (MCU or Application Processor, AP) is the core of the breast pump control system, running event detection algorithms and snapshot management logic. It receives various data from the sensor modules in real time and accurately determines the occurrence of physiological events such as milk let-down using built-in algorithms. Simultaneously, the MCU coordinates the work of various modules, sending corresponding control signals to the pump drive module based on detection results and user commands to precisely adjust the negative pressure and suction / deflating rhythm of the breast pump. Furthermore, the MCU also undertakes important tasks such as generating, storing, and filtering event snapshots, ensuring the efficient and stable operation of the entire system. The MCU can be simply referred to as the main controller.

[0245] The pump drive module is the actuating component of the breast pump that enables milk expression. It may include a DC pump or membrane pump, a solenoid valve or proportional valve. The main control unit outputs control signals via pulse width modulation (PWM) or pulse frequency modulation (PFM) to generate adjustable negative pressure output and simultaneously control the suction and deflation rhythm. Based on detected physiological events and user-set parameters, the main control unit sends control signals to the pump drive module, which then precisely adjusts the pump speed and valve opening / closing to generate the negative pressure and rhythm required by the user, simulating the sucking motion of an infant and improving the efficiency and comfort of milk expression.

[0246] The air passage and breast shield interface connects the pump drive module and the breast shield. It transmits negative pressure from the pump to the breast shield, creating a suitable negative pressure environment within the shield to achieve breast pumping. The design of the air passage and breast shield interface directly affects the efficiency and stability of negative pressure transmission. A reasonable air passage layout and sealing performance ensure accurate negative pressure transmission, reduce energy loss, and improve pumping effectiveness. Simultaneously, the material and shape of the interface must also consider user comfort and safety.

[0247] The sensing module, the sensing component of the breast pump control system, includes at least one of the following: an air pressure sensor (measuring the negative pressure N(t) inside the pump housing), a flow or milk volume related sensor (such as a pipeline differential pressure sensor, a weighing sensor, or an optical level sensor), a pump drive current detection circuit, and a vibration / sound sensor. These sensors collect various physical parameters during the breast pump's operation in real time, such as negative pressure, milk flow rate, and pump operating current, and transmit this data to the main control processing unit. By analyzing and processing this data, the main control processing unit can accurately determine the breast pump's operating status and the user's physiological condition, providing a basis for system control and decision-making. For example, the air pressure sensor can monitor changes in negative pressure inside the pump housing in real time, helping the main control processing unit determine whether a milk let-down reflex has occurred; the flow or milk volume related sensor can monitor the milk flow rate and total volume in real time, providing the user with accurate pumping information.

[0248] Non-volatile memory (EEPROM, Electrically Erasable Programmable Read Only Memory; or Flash memory) is used to store snapshot records. When the physiological event detection module detects physiological events such as milk let-down, the snapshot generation and storage module generates an event snapshot containing structured parameter vectors and tag data, and persistently stores these snapshots in the non-volatile memory. The non-volatile memory has the characteristic that data is not lost after power failure, enabling long-term storage of the user's breast pumping data and event snapshots. This data can provide a basis for subsequent snapshot filtering and matching and parameter loading, enabling personalized self-learning of breast pump parameters and providing users with a more comfortable and efficient breast pumping experience.

[0249] The human-computer interaction (HCI) module is the interface through which the user interacts with the breast pump control system. It includes a display screen or LEDs, buttons or a touchscreen, and a buzzer for status alerts and user confirmation. The display screen or LEDs can show information such as the breast pump's operating status, negative pressure, and milk flow, allowing the user to monitor the pumping process in real time. The buttons or touchscreen facilitate parameter settings and mode switching. The buzzer provides audible alerts to the user regarding the breast pump's operating status and any abnormalities. The design of the HCI module must consider user habits and ease of use, ensuring that the user can interact with the breast pump easily and accurately.

[0250] A communication module (Bluetooth BLE and / or Wi-Fi, optional) is used to synchronize user accounts and encrypted snapshots with the mobile app. Through this module, the breast pump can connect to mobile devices such as smartphones and tablets, enabling data transmission and sharing. Users can view pumping data, set pump parameters, and receive system notifications through the mobile app. Simultaneously, the communication module can synchronize encrypted event snapshots to the mobile app, facilitating data backup and management. The presence of this communication module makes the use of the breast pump more intelligent and convenient, providing a better user experience.

[0251] The mobile app (optional module) provides users with richer and more convenient functions. It offers features such as user identity management, snapshot visualization, and privacy controls. Users can register and log in to their accounts through the app to manage their personal information and breast pumping data. The snapshot visualization function displays event snapshots in intuitive charts and reports, allowing users to clearly understand their breast pumping progress and physiological characteristics. The privacy control function protects users' personal data security, allowing them to independently set the scope and permissions for data sharing. The combination of the mobile app and the breast pump makes the use of the breast pump more personalized and intelligent, meeting diverse user needs.

[0252] Reference Figure 4 The diagram illustrates the hardware module connections of the aforementioned system. The main control processing unit is electrically connected to the pump drive module, sensor module, non-volatile memory, human-machine interface module, and communication module. The negative pressure output of the pump drive module is connected to the breast pump shield via an air path and shield interface. The sensor module is positioned at the air path or shield location to collect real-time operating parameters. The communication module then establishes a wireless communication connection with an optional mobile terminal APP to achieve data synchronization and interaction. These modules cooperate to complete the adaptive parameter control process of the breast pump, providing users with a personalized breast pumping experience.

[0253] Reference Figure 5 Based on the above-mentioned system hardware modules, the control method of this disclosure may further include the following steps: Step S1, Session Initialization. When the user powers on the device or confirms "Start Breast Pumping" in the APP, the main controller (main control processing unit) establishes the Session ID for this session, loads the set of historical snapshots associated with the current user's UserID from storage (non-volatile memory), and reads the hardware capability boundaries of the machine (maximum negative pressure N_max, maximum pump frequency f_max, minimum duty cycle D_min, etc.). If a valid preferred snapshot P* exists, the automatic matching process in step S5 is initiated (executed in parallel with or before steps S2 to S4).

[0254] Step S2, Acquisition of Milk Initiation / Milk Spin Events. During breast pumping, the main controller acquires the occurrence signal of a milk initiation or milk splash event through the sensor module and / or user input. The source of the event signal may be, but is not limited to: automatic sensor detection (such as triggering by changes in signals like air pressure, flow rate, pump current, etc.), user active marking of "milk initiation" via buttons on the device or the APP, or a combination of both. The specific event detection algorithm does not constitute a limitation of this invention; the core of this invention lies in the linkage processing of mode parameters after the event occurs. The event signal at least includes the event type EventType (milk initiation or milk splash) and an optional confidence score.

[0255] Step S3: Snapshot acquisition and linkage of lactation mode parameters. This step is the core of the invention. After the event signal is output in step S2, the main controller immediately reads the relevant parameters of the lactation mode currently running by the breast pump at the event anchor time t_event, and structures them into a parameter vector P. P includes at least the following fields (see details). Figure 6 (This shows the snapshot data structure fields): •Mode_state: The current mode is either stimulation mode, sucking mode, mixed mode, or custom program, indicated by an enumeration value or program number; • Target negative pressure setting N_set: can be represented by an absolute value in kPa or a gear index, that is, the negative pressure setting that the mother is using when milk comes in / milk let-down occurs; • Pump drive frequency f_pwm: Unit Hz or cycles / minute; • The inhalation duty cycle D_inhale and the exhalation duty cycle D_exhale control the ratio of inhalation to exhalation time within each inhalation cycle; • Optional fields: Biphase second negative pressure step N2 and its duration, pressure relief strategy (fast relief / slow relief), massage module vibration intensity M_amp and frequency M_f; • Context fields: shield size code Size_code, left / right side identifier Side∈{L,R,BOTH}, offset of the event occurrence time relative to the session start point t_rel.

[0256] The parameter vector P described above fully describes the "lactation working state of the entire machine at the moment of milk initiation / milk let-down," serving as a structured linkage bridge from "event" to "mode." Unlike existing technologies that record the entire session curve or only the final state, this snapshot is precisely anchored at the moment of the physiological event.

[0257] Optionally, the negative pressure and / or flow waveform segments F before and after t_event (e.g., 5 to 30 seconds) are recorded simultaneously for subsequent similarity comparison and abnormal snapshot removal.

[0258] Step S4, Snapshot Persistence and Optimal Update. Snapshot records R=(schema_ver,UserID,HWID,fw_ver,SessionID,t_event,EventType,Side,P,F_hash,Score,milk_gain,crc) are written to the snapshot partition in storage (isolated from the firmware OTA partition), or uploaded to the APP / cloud via the communication module with encryption. A maximum of K records (K=3~10) are retained within the same (UserID,Side,EventType) bucket, and sorted according to the weighted score W=λ1·Score+λ2·normalize(milk_gain)+λ3·user_rating; or the weighted centroid P* is calculated: a score-weighted average is applied to N_set, f_pwm, and D_inhale, and then mapped to valid discrete levels.

[0259] Step S5, refer to Figure 7 The next session will automatically match and smoothly transition, achieving the core closed loop of intelligent lactation. The main controller will execute the following process upon the user's next power-on: ① Read the preferred snapshot P* bound to the current UserID and Side; if there are multiple snapshots (e.g., distinguished by time period: morning / noon / evening), perform secondary filtering by system clock or user selection.

[0260] ② Hardware boundary clipping: N_set←min(N_set,N_max_hw), f_pwm←clamp(f_pwm,f_min,f_max).

[0261] ③ Smooth transition: During the safety initialization phase (e.g., the first 5-15 seconds), start with the default low negative pressure and low frequency, and then approach the target value of P* in a piecewise linear or S-curve manner. The total transition time can be set to T_ramp (e.g., 10-30 seconds) to avoid discomfort caused by sudden changes in negative pressure.

[0262] ④ User prompt: The system will display "Your preferred mode has been loaded" via the screen or LED + buzzer. Users can undo and return to the default program with one click.

[0263] ⑤ If the user manually adjusts parameters significantly during this session, the master controller will prompt "Do you want to update the current parameters to the new snapshot?" and branch writing will be possible.

[0264] Step S6, Conflict and Security Handling.

[0265] When firmware upgrades cause program number reordering, the mapping table old_id→new_id is maintained; if it is missing, it reverts to the closest default curve family.

[0266] When changing the pump body causes a change in N_max_hw, the historical N_set is scaled proportionally and requantized to the gear.

[0267] When a user requests the deletion of private data, clear all snapshots in storage and notify the app / cloud to delete them synchronously.

[0268] Furthermore, basic linkage allows for event-triggered snapshot collection and one-click restoration for the next session.

[0269] The user manually adjusts the breast pump to a comfortable combination of parameters (e.g., suction mode, negative pressure level 5, pump frequency 60 times / minute, duty cycle 55%). During pumping, the system detects a milk let-down event (or the user presses the "milk comes in" button to confirm). At this time, the main controller 10 immediately executes the following linked operations: ① Freeze and read the current lactation mode parameter vector P = {Mode_state=sucking, N_set=gear 5, f_pwm=60, D_inhale=55%, Side=R}; ② Record the event anchor point time t_event (offset relative to the start of the current session) and the event type EventType=milk array; ③ After structuring (P, t_event, EventType, UserID, HWID), write it to storage 50 to form an "event-pattern snapshot".

[0270] The next time the user powers on, the main controller 10 detects a snapshot of the user in storage 50 and automatically sets the lactation program to the mode and parameters recorded in the snapshot (suckling mode, level 5, 60 times / minute, 55% duty cycle), instead of starting from the default stimulation mode level 1. The human-computer interaction module 60 displays "Your lactation preferences have been loaded," and the user can confirm or undo to return to the default program with one click. This way, the user does not need to manually adjust the settings each time, but can directly start pumping from the most efficient working point of the last pumping session.

[0271] Furthermore, a smooth transitional automatic matching—segmented approximation of snapshot parameters.

[0272] Based on Example 1, to avoid discomfort caused by starting with a high negative pressure immediately upon power-on, the main controller 10 performs a segmented smooth transition on the snapshot parameters: Phase 1 (0-5 seconds): Starts with the default stimulation mode of low negative pressure (level 1) and high pump frequency (120 times / minute) to give the breasts initial adaptation; Second phase (5-15 seconds): The pump frequency is linearly reduced from 120 times / min to the snapshot target of 60 times / min, and the mode is switched from stimulation to sucking; The third stage (15-25 seconds): The negative pressure increases linearly from level 1 to the target level 5 of the snapshot, and the duty cycle is adjusted to 55% simultaneously; Phase 4 (after 25 seconds): Stabilize at all parameters of snapshot P and enter normal breast pumping.

[0273] The transition curve can be linear, S-curve (smoothstep), or a non-linear strategy of first increasing frequency and then increasing voltage, which is determined by product comfort testing. If the user manually adjusts parameters during the transition (such as reducing negative pressure), the main controller will stop the automatic transition and switch to manual mode.

[0274] Furthermore, multiple snapshots are used to select the best one—accumulating multiple events to form an individual lactation profile.

[0275] As users use the storage multiple times, multiple snapshot records of the same user accumulate in storage 50 (e.g., 5 milk array snapshots). The parameters of each snapshot may differ slightly (because the parameters differ when the event is triggered after the user manually adjusts the archive in different sessions). The controller 10 executes an optimization strategy for the snapshot set: Strategy A (Latest First): Directly uses the parameters from the most recent snapshot as P*, suitable for the early breastfeeding stage where lactation status changes rapidly.

[0276] Strategy B Weighted Centroid: A weighted average is calculated for N_set, f_pwm, and D_inhale for each snapshot. The weights are determined by the snapshot's confidence score and / or the milk collection gain for that session. The calculated centroid parameter P* is then mapped to the nearest valid level. This strategy is suitable for users with relatively stable lactation status.

[0277] Strategy C User Star: Users manually mark a breast pumping experience as "this time was the best" on the APP 80, and the marked snapshot is directly used as P*.

[0278] The above three strategies can be switched in the APP settings, or the main controller can automatically select the strategy based on the number of snapshots (e.g., use strategy A if there are fewer than 3 snapshots, and switch to strategy B if there are ≥3 snapshots).

[0279] Reference Figure 8 The diagram shows a schematic of the event detection state machine, specifically a finite state machine logic diagram for milk ejection / milk release event recognition. This is the core software logic for accurately determining physiological milk ejection events and avoiding false triggering of snapshot collection. It solves the problem that simple single-threshold judgment can easily misidentify air leakage, breast shaking, and pump vibration as milk ejection. Only when the state machine flows to the confirmation (write snapshot) state will the core innovative action of collecting current lactation parameters and generating an event snapshot be executed, directly supporting the snapshot collection process of S3 and S4 mentioned above.

[0280] The state machine disclosed herein can be set to four main states, each of which undertakes a specific function: First, there's the initial, normal resting state, which is the default base state throughout the entire breast pumping process. In this state, the main controller continuously collects multiple signals from the sensor module, including negative pressure fluctuations, flow rate, and pump current, and calculates the event characteristic value J(t) in real time. This characteristic value corresponds to... Figure 7 The multi-source fusion confidence score is calculated by cyclically sampling the milk array feature index. If the feature value J(t) is greater than the threshold 1, the suspected trigger condition is met and the process jumps to the suspected state; otherwise, it remains in the resting state.

[0281] Secondly, there is a suspected state for anti-shake pre-verification. In this stage, after the signal initially exceeds the standard, the system does not immediately determine the milk flow. Instead, it conducts continuous multi-frame anti-shake verification to filter out instantaneous interference such as air leakage in the shield, body shaking, and sudden changes in negative pressure. It continuously performs multi-cycle signal acquisition for M consecutive frames and continuously verifies whether the characteristic value is stably exceeding the standard. If the signal falls back in multiple frames and the conditions are not met, it jumps back to the resting state and is judged as interference noise. No snapshot operation is performed. If all M consecutive frames meet the threshold conditions, it is determined that milk has actually come / flowed, and it enters the confirmation (write snapshot) state.

[0282] The third is the confirmation (snapshot writing) state, which is the core action node of the patent. In this state, the system officially determines that the physiological event of milk let-down / milk production has actually occurred. It will complete the core operations in sequence: anchor the event time t_event, read the complete set of lactation parameter vectors P of current negative pressure, frequency, and duty cycle, generate a structured snapshot record R and write it to non-volatile memory for persistence. At the same time, the human-computer interaction prompts the user "milk let-down detected". After the operation is completed, it automatically switches to the cooling state to prevent the snapshot from being triggered repeatedly in a short period of time.

[0283] Finally, there is the cooling state, which serves as a lockout period to prevent repeated triggering. Since the let-down reflex is a physiological process that lasts for several minutes, not setting a cooling window will cause the continuous milk flow to repeatedly trigger events and generate a large number of redundant snapshots. Therefore, a cooling timer T_cool is set for this state. During the lockout period, the let-down reflex event detection logic is blocked, and no new events are judged. After the cooling time T_cool expires, it automatically returns to the resting state and restarts a new round of let-down reflex detection.

[0284] This state machine forms a complete closed-loop state transition, containing two core links: The first link is for invalid events caused by interference and noise. The transition path is: rest → J(t) exceeds threshold 1 → suspected → multiple frames fail verification → return to rest. Its function is to filter out false signals such as instantaneous negative pressure fluctuations, air leakage, and jitter, and to prevent invalid snapshots from polluting the storage. The second link is for valid events caused by actual milk flow, which is also the core process of the patent. The transition path is: rest → J(t) > threshold 1 → suspected → consecutive M frames of stable compliance → confirmation (write snapshot) → cooling timer → T_cool timeout → return to rest. This completely corresponds to the patent process S2 event acquisition, S3 snapshot acquisition, and S4 snapshot persistence.

[0285] The aforementioned state machine addresses several shortcomings of existing technologies: First, it resolves the issue of misjudgment in real-time detection in existing solutions such as Willow and Medela. Existing solutions only perform single-threshold judgments, and even slight air leakage or body movement can lead to misjudgment of milk flow, causing frequent erroneous mode switching. This state machine adds dual verification of "suspected multi-frame stabilization + cooling lock-in period," significantly reducing the probability of false snapshots. Second, it supports a snapshot confidence scoring mechanism; achieving stable multi-frame performance increases the event confidence score. Figure 7 The evaluation metrics include: low-confidence interference signals are directly discarded and not written to the snapshot, which is an auxiliary support for the optimization mechanism of the corresponding document's fifth drawback; thirdly, it reduces the MCU computing power overhead, only performs lightweight sampling calculations in the resting state, and executes complex fusion algorithms only in the suspected / confirmed stage, without the need for high-load closed-loop adjustment throughout the process, and is compatible with the low-cost main control chip of breast pumps; fourthly, it avoids the generation of multiple duplicate snapshots for the same milk ejection, and prevents repeated snapshot triggering during the continuous secretion of a single milk ejection during the cooling window period, ensuring that each snapshot corresponds to an independent physiological node of milk ejection, making the snapshot library data cleaner, and the subsequent selection strategy more accurate.

[0286] In some examples, users manually trigger the let-down signal by pressing a physical button on the breast pump or by operating the terminal software.

[0287] The solution of triggering the milk let-down signal through physical buttons does not require additional modification to the hardware structure of the breast pump. It can be achieved using the operation buttons already configured on the breast pump itself. The modification cost is low, it is compatible with most existing mass-produced breast pump products, the operation is intuitive and convenient for users, and the marking can be completed quickly without the need to connect to an additional smart terminal. It is suitable for breast pumping scenarios without network or without carrying mobile devices. The solution of triggering milk let-down signal through terminal software allows users to free their hands during breast pumping, without having to reach out to operate the buttons on the breast pump itself to complete the marking. At the same time, the terminal can synchronously store and organize data on the occurrence time and duration of milk let-downs during multiple pumping sessions, making it convenient for users to track their own lactation patterns over the long term, and also providing complete data support for medical staff to assess lactation status.

[0288] The two triggering methods can meet the needs of different users and different scenarios, greatly improving the adaptability of the breast pump parameter adjustment. This allows the negative pressure and frequency adjustment of the breast pump to better match the user's own lactation rhythm, which can not only improve the efficiency of milk expression and reduce the time spent on a single milk expression, but also avoid the pressure damage to the nipple and areola caused by mismatched parameters, thus improving the comfort of the milk expression process.

[0289] The user's manual triggering of the milk let-down marker can accurately correspond to the user's own lactation experience, avoiding misjudgment or missed judgment in the intelligent milk let-down recognition, improving the accuracy of subsequent parameter adjustment logic, and making the breast pump's parameter control more in line with the user's actual lactation needs. Long-term use can help maintain a stable lactation state and improve the success rate of breastfeeding.

[0290] In some examples, weighted centroid filtering strategies include: For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk volume gain parameter and the parameter confidence parameter of that snapshot. The weighted sum of each parameter of all snapshots that meet the criteria is then divided by the sum of all weights to obtain the weighted centroid parameter vector. The target event snapshot is constructed using the weighted centroid parameter vector.

[0291] The weight of a single event snapshot is calculated by multiplying the milk yield gain parameter and the parameter confidence parameter. This approach can simultaneously take into account the actual milk yield improvement effect during the milk pumping process and the reliability of the current parameter combination. It avoids the bias caused by selecting parameters based solely on milk yield or model confidence. It will not ignore high-potential effective parameter combinations under low milk yield gain, nor will it exclude personalized parameters with low confidence but actually suitable for the current milk yield status, thus improving the scientific nature of parameter selection. The weighted centroid parameter vector is obtained by taking a weighted average of all snapshot parameters that meet the search criteria. This can integrate the parameter features of multiple effective breast pumping processes, eliminate parameter errors caused by random lactation fluctuations in a single event snapshot, and make the final target parameters more consistent with the user's current actual breast condition and lactation pattern. Compared with the method of directly selecting a single optimal snapshot, the parameter stability is stronger. By constructing a target event snapshot using a weighted centroid parameter vector as the basis for parameter adjustment, personalized dynamic adaptation of breast pump parameters can be achieved. This can adapt to individual differences in milk production volume and rhythm among different users, as well as to the changing needs of the same user at different lactation stages and in different pumping scenarios. This effectively improves the comfort of the pumping process and maximizes the milk production volume per pumping session, solving the problems of poor adaptability of existing fixed parameter modes and unstable experience of manual parameter adjustment.

[0292] In some examples, a program module for performing the breast pump control method described above is also included.

[0293] By integrating the program module into the control unit of the breast pump, modular deployment and iterative updates of the control logic can be achieved.

[0294] Modular design reduces the coupling between different functional modules, allowing R&D personnel to develop and debug independent functions such as lactation mode adjustment, negative pressure calibration, and user preference adaptation separately without modifying the overall control framework. This significantly shortens the development cycle of new breast pump products and reduces the cost of later function upgrades. Manufacturers can directly update the control logic of sold products online via Over-the-Air (OTA) technology without requiring users to replace hardware. On the other hand, the modular structure improves the operational stability of the control program. If a single functional module malfunctions, the fault node can be quickly located without triggering a crash in the entire control program. It also facilitates after-sales maintenance personnel to quickly troubleshoot program-related issues, reducing the difficulty of after-sales maintenance.

[0295] The program module can interface with the breast pump's built-in storage unit and Bluetooth communication module, supporting automatic recording of the user's parameter adjustment habits during each use. Based on multiple usage data, it generates personalized parameter recommendations tailored to the user's lactation patterns. Users no longer need to manually readjust the negative pressure and suction frequency each time they use the pump; they can directly call the adapted parameters to start pumping, significantly improving the user experience. At the same time, the program module can also automatically monitor abnormal parameters: when the breast pump experiences abnormal increases / decreases in negative pressure or motor operating parameters deviating from the safe range, the program module can quickly trigger the protection mechanism, automatically reducing power or shutting down the pump, and pushing an abnormality reminder to the user's linked mobile phone. This not only prevents abnormal parameters from damaging the user's mammary glands but also promptly reminds the user to maintain the equipment, extending the lifespan of the breast pump.

[0296] For smart breast pumps with multiple speeds and modes, the program module can achieve smooth parameter switching between different modes, avoiding the sudden change in negative pressure that can stimulate the user's mammary glands caused by traditional hard switching methods, making the breast pumping process more comfortable and natural. At the same time, the standardized program module interface is also compatible with different hardware configurations. Manufacturers only need to adjust the parameter configuration file to adapt to breast pump products with different positioning and power, without having to redevelop the entire control program, which further improves the efficiency of product development and adaptation and reduces the development costs of multiple product lines.

[0297] In some examples, a cloud server is also included, which stores snapshot data of the group of users and is used to send the group's prior parameter vectors corresponding to the model and nipple size to the breast pump upon request. The group prior parameter vector is the cluster centroid parameter obtained by clustering effective event snapshots of users of the same model and size in the cloud.

[0298] By storing snapshot data of aggregated users on cloud servers, clustering calculations can be performed based on massive amounts of real usage data from users with the same breast pump model and nipple size. This yields a group prior parameter vector that better reflects the actual usage needs of most users. Compared to individual users having to figure out and adjust parameters from scratch, users can directly download initial parameters that match their own situation after unpacking the device. This significantly shortens the parameter adaptation cycle for novice users, reduces the operational costs of manual adjustments, avoids problems such as breast pumping pain and low lactation efficiency caused by unreasonable initial parameter settings, and improves the initial user experience.

[0299] Based on cluster centroid parameters generated from group user data, the parameter benchmark can be optimized by comprehensively considering the usage habits of different users, avoiding parameter deviations caused by individual user errors, and significantly improving the rationality of the initial parameters. At the same time, the prior parameters obtained by cluster calculation retain the effective lactation pattern characteristics of users of the same type, which can provide a more scientific starting benchmark for users' subsequent personalized parameter adjustments, accelerate the convergence speed of users' personal adaptation parameters, and find the most suitable lactation parameter combination for themselves more quickly.

[0300] The cloud-based dynamic updating of user data can continuously optimize the prior parameter vector. As the amount of user data of the same model and size continues to accumulate, the centroid parameters obtained by clustering calculation will continuously match the actual usage needs. New users can always obtain the current optimal initial parameters. The parameter scheme can be iterated and optimized without the need for manufacturers to upgrade hardware firmware, which reduces the product's operation and maintenance costs. At the same time, it allows existing users to obtain the updated optimized parameters at any time, providing users with a continuously upgraded user experience.

[0301] The aforementioned parameter scheme also enables the reuse of user data as a group value. By clustering and mining the common patterns in massive user data in the cloud, the actual usage behavior of ordinary users is transformed into parameter experience that can be shared by the entire group. This eliminates the need for professionals to manually complete the customized development of parameter schemes and allows parameter schemes to automatically adapt to the needs of different user subgroups, improving the precision of breast pump parameter adaptation and reducing the development cost of product parameter schemes.

[0302] This disclosure also has the following beneficial effects: Snapshot capture is triggered by actual milk ejection / let-off events, rather than switching at fixed intervals. This ensures that the recorded parameter combinations truly correspond to the most efficient moments of milk ejection, resulting in high accuracy when reproducing the event across sessions.

[0303] Snapshots are persistently stored in non-volatile memory, turning the next breast pumping session from a "cold start" to a "warm start." Even without relying on a continuous closed-loop system, historical optimal parameter combinations can be reproduced.

[0304] The snapshot record contains the complete parameter vector, event anchor time, and confidence level. During the next match, only this information needs to be read and cropped to the hardware's boundary, achieving decoupling and closed-loop processing of "real-time detection" and "cross-session reproduction."

[0305] Side-by-side / time-by-time bucketing prevents parameter misuse; hardware trimming, smooth transition, and one-click undo functions ensure safety and comfort during use; firmware mapping ensures upgrade compatibility; and update prompts after user parameter tuning ensure that snapshots always reflect the user's latest preferences.

[0306] Additional benefits: The main incremental improvements of this disclosure are in software and a small amount of storage overhead (64 bytes / record × 200 records ≈ 12.5KB). It can be deployed to existing product lines via OTA firmware upgrades without increasing hardware bill of materials (BOM) costs.

[0307] On the other hand, this disclosure also provides a breast pump, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the parameter control method of the breast pump described above.

[0308] The aforementioned breast pump, through preset parameter control logic, compared to the traditional breast pump's control scheme which uses a fixed lactation mode, fixed vacuum pressure level, and fixed pumping rhythm, can automatically and dynamically adapt the pressure and sucking frequency according to the individual user's mammary gland condition, lactation pattern, and pain tolerance. It will not cause nipple and areola damage or breast engorgement due to excessive pressure, nor will it cause incomplete lactation or low milk expression efficiency due to insufficient pressure or mismatched rhythm, effectively improving the comfort and thoroughness of the milk expression process. For users who are mixing breastfeeding or trying to increase milk supply, this breast pump can adjust its parameter curves based on milk production data collected from multiple pumping sessions. This gradually stimulates the mammary glands to increase milk production, helping users establish a milk production pattern that suits their individual needs and improving the success rate of increasing milk supply. At the same time, this solution achieves adaptive parameter adjustment through software algorithms, eliminating the need for additional complex mechanical adjustment structures on the breast pump itself. This reduces the hardware production cost and the probability of structural failure, while also simplifying the user's operation process. Users can obtain a suitable pumping mode without manually and repeatedly adjusting the settings, lowering the barrier to entry and improving the user experience. The above parameter control process can store user usage habit data in real time. When the user uses the breast pump again, the adapted personalized parameters can be directly called up without repeated adaptation, which further optimizes the convenience of continuous use. At the same time, it adapts to the body changes at different stages of lactation, and can provide adapted parameter schemes from the early stage of lactation to the supply and demand balance stage, covering the user's usage needs throughout the entire breastfeeding cycle.

[0309] Furthermore, in the factory settings disclosed herein, the breast pump automatically switches from stimulation mode to lactation mode by default during the T1 working time. The default T1 duration is typically 1-5 minutes, with the specific duration optimized for different product models and user groups.

[0310] In stimulation mode, the breast pump operates at a higher frequency and lower negative pressure, simulating the actions of a baby when they first start suckling, to help mothers stimulate milk production. The lactation mode uses a lower frequency and higher negative pressure to simulate the actions of a baby suckling effectively, improving milk extraction efficiency. During the pumping process, the system uses sensors to monitor milk flow, pressure, and other data in real time to detect letdown time.

[0311] When the system detects that milk flow has reached a certain threshold or that there is a significant change in pressure, it determines that letdown has occurred. The criteria for determining letdown time are adjusted based on different products and users to ensure accuracy. Then, during subsequent pumping sessions, the system adaptively updates T1 based on historical letdown times.

[0312] The specific algorithm works by recording the letdown time of each pumping session and analyzing and processing this data.

[0313] By calculating statistical indicators such as the average and median of historical letdown times, the system can determine a reasonable T1 duration. At the same time, the system will also consider user habits and feedback to further optimize the T1 duration.

[0314] For example, if a user frequently adjusts the breast pump's mode manually after a letdown, the system will automatically adjust the T1 duration based on these operation records to better suit the user's needs. This allows for personalized customization based on the new automatic T1 switching mode. Users can also manually adjust the T1 duration according to their needs, or choose to disable the automatic switching function and manually control the breast pump's mode switching. Furthermore, the system will continuously optimize the T1 duration and mode switching strategy based on the user's usage habits and physical condition, providing a more personalized breast pumping experience.

[0315] On the other hand, this disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0316] By solidifying the parameter control logic into an independently callable computer program and storing it in a readable storage medium, the breast pump parameter control method can be quickly ported and deployed across devices. Whether it is used to upgrade the firmware of existing breast pump products or integrated into newly developed smart breast pump hardware, there is no need to redevelop the core control logic, which greatly reduces the R&D iteration cost of smart breast pumps and shortens the product launch cycle. The standardized program storage format facilitates subsequent updates to parameter logic based on lactation data and user feedback from different user groups. Users can obtain a control scheme that is more suitable for their own lactation patterns simply by upgrading online via OTA, and can continue to enjoy a better user experience without replacing hardware. The independently stored control program also facilitates function debugging and problem tracing. R&D personnel can quickly locate anomalies at the control logic level, improve product maintenance efficiency and stability, and thus ensure the reliability of the breast pump's lactation mode adjustment, providing users with a continuous, stable, and comfortable breast pumping experience.

[0317] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0318] In summary, after reading this detailed disclosure, those skilled in the art will understand that the foregoing detailed disclosure may be presented by way of example only and may not be restrictive. Although not explicitly stated herein, those skilled in the art will understand that the requirements of this disclosure encompass various reasonable changes, improvements, and modifications to the embodiments. These changes, improvements, and modifications are intended to be made by this disclosure and are within the spirit and scope of the exemplary embodiments of this disclosure.

[0319] Furthermore, certain terms used in this disclosure have been used to describe embodiments of this disclosure. For example, "an embodiment," "an embodiment," and / or "some embodiments" mean that a particular feature, structure, or characteristic described in connection with that embodiment may be included in at least one embodiment of this disclosure. Therefore, it is to be emphasized and understood that two or more references to "an embodiment" or "an embodiment" or "alternative embodiment" in various parts of this disclosure do not necessarily refer to the same embodiment. Moreover, specific features, structures, or characteristics may be suitably combined in one or more embodiments of this disclosure.

[0320] It should be understood that in the foregoing description of the embodiments of this disclosure, various features are combined in a single embodiment, drawing, or description for the purpose of simplifying the disclosure and to aid in understanding a feature. However, this does not mean that the combination of these features is necessary, and those skilled in the art, upon reading this disclosure, may readily identify some of the devices as separate embodiments. That is, the embodiments in this disclosure can also be understood as an integration of multiple secondary embodiments. It is also valid when each secondary embodiment contains fewer than all the features of a single foregoing disclosed embodiment.

[0321] Every patent, patent publication, publication of a patent publication, and other material, such as articles, books, specifications, publications, documents, and literature (excluding any related historical examination documents), cited in this disclosure is incorporated herein for all purposes, including, for example, in the specification and claims of this disclosure. However, in the event of any inconsistency or conflict between the descriptions, definitions, and / or terms used in the foregoing and those used in this disclosure, the descriptions, definitions, and / or terms used in this disclosure shall prevail.

[0322] Finally, it should be understood that the disclosed embodiments herein are illustrative of the principles of the embodiments of this disclosure. Other modified embodiments are also within the scope of this disclosure. Therefore, the embodiments disclosed herein are merely examples and not limitations. Those skilled in the art can implement the disclosures herein by using alternative configurations based on the embodiments in this disclosure. Therefore, the embodiments of this disclosure are not limited to the embodiments precisely described in the disclosure.

Claims

1. A parameter control method for a breast pump, characterized in that, include: Detect milk ejection physiological events to determine whether a milk-inducing event or a milk ejection event has occurred. When a milk initiation event or a milk let-down event occurs, the current operating parameters of the breast pump are captured, and an event snapshot with added tagged data and a structured parameter vector is generated. The structured parameter vector includes at least the vacuum parameter, the milk pumping frequency parameter, and the suction duty cycle parameter. The event snapshot is persistently stored in the snapshot storage collection; In response to the breast pump start request, the target event snapshot is obtained by matching from the snapshot storage set according to the preset filtering strategy; The parameter vector of the target event snapshot is loaded, and after version mapping and hardware adaptation adjustments, the current operating parameters of the breast pump are obtained.

2. The parameter control method for a breast pump according to claim 1, characterized in that, The detected lactation physiological events include: The negative pressure value of the breast pump is continuously collected by a pressure sensor; Calculate the change in negative pressure per unit time; When the change in negative pressure exceeds a preset threshold, a milk inflow event or a milk flow event is determined to have occurred.

3. The parameter control method for a breast pump according to claim 1, characterized in that, The detected lactation physiological events include: It receives milk ejection signal signals manually triggered by the user through physical buttons on the breast pump or by operating the terminal software, and determines whether a milk initiation event or a milk ejection event has occurred.

4. The parameter control method for a breast pump according to claim 1, characterized in that, The tag data for the event snapshot includes at least a breastfeeding side identifier and a collection period tag; The breast pump side indicator is automatically identified by a sensor or manually selected by the user. The collection period labels are automatically generated based on the current time.

5. The parameter control method for a breast pump according to claim 1, characterized in that, The step of matching the target event snapshot from the snapshot storage set according to the preset filtering strategy includes: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than a preset number threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots that meet the retrieval criteria in the snapshot storage set is greater than or equal to a preset threshold, a weighted centroid filtering strategy is used to generate a target event snapshot.

6. The parameter control method for a breast pump according to claim 1, characterized in that, The preset filtering strategy includes: When the number of event snapshots that meet the search criteria in the snapshot storage set is less than a preset threshold, a time-series filtering strategy is adopted to directly select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots in the snapshot storage set that meet the search criteria is greater than or equal to the preset threshold, the system automatically switches to a weighted centroid filtering strategy. For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot. For all snapshots that meet the conditions, each parameter is summed with weights and then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is then constructed using the weighted centroid parameter vector.

7. The parameter control method for a breast pump according to claim 6, characterized in that, The weighted centroid selection strategy includes: For each event snapshot that meets the search criteria, the weight is calculated as the product of the milk yield gain parameter and the parameter confidence parameter of that snapshot; For all snapshots that meet the conditions, each parameter is weighted and summed, and then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is constructed using the weighted centroid parameter vector.

8. The parameter control method for a breast pump according to claim 6, characterized in that, The preset filtering strategy also includes a user star filtering strategy. When there are event snapshots in the snapshot storage set that meet the search conditions and are marked by the user star, the event snapshot marked by the user star is selected as the target event snapshot.

9. The parameter control method for a breast pump according to claim 1, characterized in that, The process of loading the parameter vector of the target event snapshot includes: First, read the firmware version of the current device and the model of the currently connected pump body, and then compare them with the original firmware version and the original pump body model recorded in the target event snapshot. If there are version differences, the original parameter number is mapped to the parameter number corresponding to the current firmware version through a pre-stored version mapping table.

10. The parameter control method for a breast pump according to claim 9, characterized in that, The steps after obtaining the mapped parameters also include: The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump hardware to obtain a parameter vector adapted to the current hardware, which serves as the current operating parameters of the breast pump.

11. The parameter control method for a breast pump according to claim 1, characterized in that, The structured parameter vector of the event snapshot includes at least: vacuum degree parameter, milk suction frequency parameter, and suction duty cycle parameter.

12. The parameter control method for a breast pump according to claim 1, characterized in that, The event snapshot includes at least one of the following tag data: The data includes the following information: milk pumping side identifier, data collection period label, firmware version number at the time of data collection, pump model identifier at the time of data collection, user star mark, milk volume gain parameter at the time of data collection, and parameter confidence level parameter.

13. The parameter control method for a breast pump according to claim 1, characterized in that, Before the step of responding to the breast pump start request and matching the target event snapshot from the snapshot storage set according to a preset filtering strategy, the method further includes a side-by-side and time-based bucket retrieval step: Obtain the milk pumping side identifier and the time period to which the current milk pumping request belongs, and match event snapshots from the snapshot storage set that have the same milk pumping side identifier and the same time period label to participate in the filtering.

14. The parameter control method for a breast pump according to claim 13, characterized in that, When no event snapshot matches the current breast pumping side identifier, the independent search for each side is not triggered, and the global snapshot is used for filtering. If no event snapshot matches the time period label of the current time, time period matching is not triggered, and all snapshots of the corresponding breast pumping side are used for filtering.

15. The parameter control method for a breast pump according to claim 1, characterized in that, It also includes cross-version compatibility mapping steps: When loading the parameters of the event snapshot, read the firmware version of the current device and the model of the currently connected pump body, and compare them with the original firmware version and the original pump body model recorded in the event snapshot. If there are version differences, the original parameter number will be mapped to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. The snapshot parameters are cropped and scaled according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

16. The parameter control method for a breast pump according to claim 15, characterized in that, The clipping of the parameter boundaries includes: If the vacuum parameter in the event snapshot is higher than the current negative pressure safety limit preset by the pump body hardware, the vacuum parameter will be automatically adjusted to the negative pressure safety limit, while retaining the proportional relationship of the other parameters.

17. The parameter control method for a breast pump according to any one of claims 1 to 16, characterized in that, It also includes the steps for controlling the milk array window linkage: When a milk ejection event is detected to enter the milk ejection duration window, a short-term enhancement adjustment is triggered, and the current operating parameters are adjusted to the enhancement parameters. When generating an event snapshot, the running parameters are extracted from the moment after the milk flow occurs and before the enhanced adjustment takes effect. The enhanced parameters after the short-term enhanced adjustment are not recorded.

18. The parameter control method for a breast pump according to any one of claims 1 to 16, characterized in that, It also includes the snapshot incremental update step: When a user manually adjusts the running parameters in the current lactation session, a confirmation prompt is output asking whether to update the snapshot set, and the user's confirmation response is received. If the user confirms the update, the adjusted parameters will be written as a new event snapshot based on the current capacity of the snapshot storage set. When the storage capacity exceeds the limit, the event snapshot generated earliest will be replaced. If the user confirms that they will not update, the adjusted parameters will be valid for the current lactation session and will not be written to the snapshot storage set.

19. The parameter control method for a breast pump according to claim 18, characterized in that, When the user confirms the update, it is determined whether the difference between the manually adjusted parameters and the target parameters in the existing snapshot exceeds a preset difference threshold. If it does not exceed the preset difference threshold, the weight parameters of the corresponding existing snapshot are updated directly without adding a new snapshot entry.

20. The parameter control method for a breast pump according to any one of claims 1 to 16, characterized in that, It also includes snapshot processing steps in special security modes: When entering the special security mode, security audits are performed on all event snapshot parameters to be loaded in the snapshot storage set. Parameters that exceed the preset parameter boundaries of the special security mode are forcibly pruned to within the boundary values. An audit flag is added to the pruned event snapshots, and the pruned parameters are loaded without discarding the original snapshot data.

21. The parameter control method for a breast pump according to claim 20, characterized in that, The special safety mode is either a nipple hypersensitivity mode or a premature infant breastfeeding mode, and the upper limit of the negative pressure of the preset parameter boundary of the special safety mode is lower than the upper limit of the negative pressure of the normal mode.

22. The parameter control method for a breast pump according to any one of claims 1 to 16, characterized in that, When no event snapshot corresponding to the current search criteria exists in the snapshot storage set, a cold start process is triggered, which includes: Obtain the group prior parameter vector of users with the same breast pump model and nipple size from the cloud, and use the group prior parameter vector as the initial recommendation parameters. After each new individual event snapshot is generated, the fusion weight of the group prior parameters is reduced; When the number of individual event snapshots reaches the preset cold start threshold, the fusion weight of the group prior parameters is adjusted to zero, and the filtering results of individual snapshots are used.

23. The parameter control method for a breast pump according to claim 22, characterized in that, The group prior parameter vector is the cluster centroid parameter obtained by clustering effective event snapshots of users of the same model and size in the cloud.

24. A breast pump control system, characterized in that, The method for performing parameter control of a breast pump according to any one of claims 1 to 23, the breast pump control system comprising a physiological event detection module, a snapshot generation and storage module, a snapshot filtering and matching module, and a parameter loading module; The physiological event detection module is used to continuously collect the negative pressure value of the breast pump through a pressure sensor, calculate the change in negative pressure per unit time, and determine the occurrence of a milk initiation event or a milk ejection event when the change in negative pressure exceeds a preset threshold, or receive a milk ejection marker signal manually triggered by the user. The snapshot generation and storage module is electrically connected to the physiological event detection module. It is used to capture the current operating parameters at the moment a physiological event occurs, generate an event snapshot containing structured parameter vectors and label data, and persistently store the event snapshot in the snapshot storage collection. The snapshot filtering and matching module is electrically connected to the snapshot generation and storage module, and is used to match the target event snapshot from the snapshot storage set according to a preset filtering strategy when a start request is received. When the number of event snapshots that meet the search criteria in the snapshot storage set is less than a preset number threshold, a time-series filtering strategy is adopted to select the event snapshot with the latest generation time that meets the search criteria as the target event snapshot. When the number of event snapshots that meet the retrieval criteria in the snapshot storage set is greater than or equal to the preset number threshold, a weighted centroid filtering strategy is used to construct the target event snapshot. The parameter loading module is used to load the parameters of the target event snapshot as the current running parameters. During loading, it first reads the firmware version of the current device and the model of the currently connected pump body hardware, compares the original firmware version and the original pump body model recorded in the event snapshot, and if there is a version difference, it maps the original parameter number to the parameter number corresponding to the current firmware version through the pre-stored firmware version mapping table. Then, it trims and scales the snapshot parameters according to the parameter boundaries of the current pump body hardware to obtain a parameter vector adapted to the current hardware.

25. The breast pump control system according to claim 24, characterized in that, The method for manually triggering the milk let-down signal is for the user to trigger it via a physical button on the breast pump or via terminal software.

26. The breast pump control system according to claim 24, characterized in that, The weighted centroid selection strategy includes: For each event snapshot that meets the search criteria, its weight is calculated as the product of the milk volume gain parameter and the parameter confidence parameter of that snapshot. The parameters of all snapshots that meet the criteria are weighted and summed separately, and then divided by the sum of all weights to obtain a weighted centroid parameter vector. The target event snapshot is constructed using the weighted centroid parameter vector.

27. The breast pump control system according to claim 24, characterized in that, It also includes a program module for performing the parameter control method of the breast pump according to any one of claims 1-18.

28. The breast pump control system according to claim 24, characterized in that, It also includes a cloud server that stores snapshot data of a group of users, which is used to send a group prior parameter vector of the corresponding model and corresponding nipple size to the breast pump according to the request; The group prior parameter vector is the cluster centroid parameter obtained by clustering effective event snapshots of users of the same model and size in the cloud.

29. A breast pump, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the parameter control method for the breast pump according to any one of claims 1 to 23.

30. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the parameter control method for the breast pump according to any one of claims 1 to 23.

Images