A stress assessment method and system based on active physical interaction

Abstract

The present application relates to the technical field of stress evaluation, and in particular to a stress evaluation method and system based on active physical interaction, which comprises: obtaining passive physiological data, the passive physiological data comprising at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data and sleep quality data; calculating a physiological load level according to the passive physiological data; updating a physiological load baseline according to scenario data; the physiological load baseline being a dynamic threshold for judging whether the stress level is abnormal, which changes with the scenario in which the user is located; and comprehensively determining the stress level of the user according to the physiological load baseline and the physiological load level. The present application can greatly improve the accuracy of stress evaluation of a user by a wearable device, reduce the misjudgment rate and improve the user experience.

Description

Technical Field

[0001] This invention relates to the field of stress assessment technology, and specifically to a stress assessment method and system based on active physical interaction. Background Technology

[0002] With the widespread adoption of smart wearable devices, stress monitoring based on physiological data collected by these devices has become an important area of ​​research in stress assessment.

[0003] For example, patent application CN120744837A proposes a method and system for time-series fusion of multidimensional physiological data and prediction of health status, relating to the field of electronic digital data processing technology. This invention uses high-precision sensing equipment and parameter calibration models to calibrate the data, and achieves time synchronization and error compensation through a weighted fusion algorithm, improving the accuracy of physiological parameters. Simultaneously, it utilizes a closed-loop feedback mechanism to correct fatigue and stress indices in real time, and generates a comprehensive status index based on the assessment results for further analysis and prediction. Machine learning technology is used to predict the changing trend of the comprehensive status index, and combined with the user's physiological data, targeted dynamic rehabilitation plans are provided, including training intensity optimization, dietary adjustments, and rest cycle planning. A timed feedback mechanism is established to allow users to obtain real-time changes in their health status and adjustment suggestions, thereby improving the scientific rigor and timeliness of rehabilitation plans, significantly optimizing rehabilitation effects and efficiency, and achieving intelligent health management. Currently, traditional technologies have also attempted to propose some solutions to improve the accuracy of stress assessment through multimodal data fusion.

[0004] For example, patent application CN120766975A discloses a decompression ring control method, system, device, and medium based on scene recognition. The method includes: acquiring physiological data; calculating a first data difference between the physiological data and a preset physiological data benchmark; obtaining a first index based on the first data difference; obtaining the first index from a preset first index mapping table; acquiring physical data; calculating a second data difference between the physical data and a preset physical data benchmark; obtaining a second index based on the second data difference; obtaining the second index from a preset second index mapping table; calculating a first pressure state based on the first and second indicators, the first pressure state representing the user's current pressure level; and generating a corresponding first decompression scheme based on the first pressure state.

[0005] It can be seen that the traditional approach involves calculating the difference between physiological data, physical data and a preset benchmark to obtain the first and second indicators, and then weighting and fusing the two to determine the stress state.

[0006] However, the applicant noted that these stress state calculation methods still have room for improvement in adaptability and accuracy when faced with complex and dynamically changing real-world application scenarios. Summary of the Invention

[0007] The purpose of this invention is to provide a stress assessment method and system based on active physical interaction, which partially solves or alleviates the above-mentioned shortcomings in the prior art, and can greatly improve the accuracy of wearable devices in assessing user stress, reduce the false judgment rate, and improve user experience.

[0008] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution: A first aspect of the present invention is to provide a stress assessment method based on active physical interaction, comprising: S101: Acquire passive physiological data, wherein the passive physiological data includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data; S102: Calculate the physiological load level based on the passive physiological data; S103: Update the physiological load baseline based on scenario data; the physiological load baseline refers to a dynamic threshold that changes with the user's scenario and is used to determine whether the stress level is abnormal; wherein, S103 includes: S1031: Acquire the scenario data, which includes active physical data. The active physical data refers to the rotation data generated by the wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotational angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. S1032: Generate at least one scenario intent label based on the scenario data; S1033: Update the physiological load baseline based on the scenario intent label; S104: Determine the user's stress level based on the physiological load baseline and the physiological load level.

[0009] In some embodiments, S1032 includes: A scenario confirmation signal is issued; the scenario confirmation signal is used to inquire whether the scenario intent label is accurate.

[0010] In some embodiments, S1032 further includes: Generate proactive contextual intent tags based on the proactive physical data; Passive contextual intent tags are generated based on the current time, geographical location, heart rate data, and the aforementioned sleep quality data; The context intent tag is generated based on the active context intent tag and the passive context intent tag.

[0011] In some embodiments, the rotation law refers to the change pattern reflected by at least one of the number of rotations, the rotation speed, the rotational angular acceleration, and the rotation direction; And / or, when generating the context intent tag, the weight of the active context intent tag is greater than that of the passive context intent tag.

[0012] In some embodiments, it also includes: Based on the pressure level, a corresponding decompression solution is generated, including: The pressure level is input into the decompression scheme generation model, which is used to generate the corresponding decompression scheme based on the pressure level. The decompression scheme includes guiding the user to adjust their behavior pattern through vibration of the wearable device and / or the display interface of other devices connected to the wearable device.

[0013] In some embodiments, the wearable device includes a smart ring having an outer ring and an inner ring, the outer ring and the inner ring being rotatable relative to each other.

[0014] In some embodiments, the context data further includes scenario data; the scenario data includes activity status data, and / or geographic location and / or calendar event information.

[0015] In some embodiments, it also includes: Delete the passive physiological data and / or the active physical data that are less than or equal to a preset threshold.

[0016] In some embodiments, S1032 further includes: The contextual intent label is generated based on a pre-trained contextual intent recognition model; wherein the contextual intent recognition model is trained using a historical contextual dataset, which includes: the active physical data generated by the subject wearing the wearable device, and the corresponding contextual information.

[0017] A second aspect of the present invention is to provide a stress assessment system based on active physical interaction, comprising: A passive physiological data acquisition module is used to acquire passive physiological data, which includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data. A physiological load level calculation module is used to calculate the physiological load level based on the passive physiological data; A physiological load baseline update module is used to update the physiological load baseline based on scenario data; the physiological load baseline refers to a dynamic threshold that changes with the user's scenario and is used to determine whether the stress level is abnormal; wherein, the physiological load baseline update module further includes: The scenario data acquisition unit is used to acquire the scenario data, which includes active physical data, which refers to the rotation data generated by a wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotation angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. A scenario intent tag generation unit is used to generate at least one scenario intent tag based on the scenario data; A physiological workload baseline update unit is used to update the physiological workload baseline according to the scenario intent label; The stress level determination module is used to determine the user's stress level based on the physiological load baseline and the physiological load level.

[0018] Beneficial technical effects: This invention prioritizes the relative levels of passive physiological data while using the sequence characteristics of active physical data as a reference for stress identification. For example, it uses actively input physical data as a discriminant of the user's situation, dynamically adjusting the stress level determination mechanism based on this situation (e.g., adjusting the physiological load baseline). Another example is the use of the sequence characteristics of physical data over a period of time (typically, average rotational speed or rotational patterns over a period of time) to dynamically adjust the physiological load baseline. Specifically, Firstly, this invention distinguishes scenarios through physical data, rather than directly calculating the pressure level based on the high or low levels of physical data, which improves accuracy. In other words, this scenario distinction can further improve the accuracy of pressure recognition based on the user's personalized usage habits.

[0019] Secondly, the dynamic update mechanism for physiological load baseline provided by this invention sets different stress judgment criteria for different scenarios, that is, it adapts different stress recognition sensitivities to users in different scenarios, rather than directly equating physiological load with psychological stress. This achieves a certain degree of in-depth identification of stress sources and greatly reduces misjudgment.

[0020] Third, this invention generates contextual intent tags for the current situation by integrating active and passive contextual intent tags. It can see both the user's "external situation" (passive) and hear the user's "behavioral intent" (active), which greatly improves the contextual adaptability of stress level assessment and helps to provide accurate, personalized and timely effective intervention for users. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0022] Figure 1 This is a flowchart illustrating a stress assessment method based on active physical interaction in an exemplary embodiment of the present invention. Figure 2 This is an example diagram of a stress assessment method based on active physical interaction in an exemplary embodiment of the present invention; Figure 3 This is another example diagram of a stress assessment method based on active physical interaction in an exemplary embodiment of the present invention; Figure 4 This is another flowchart illustrating the stress assessment method based on active physical interaction in an exemplary embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a stress assessment system based on active physical interaction in an exemplary embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0024] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0025] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0026] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0027] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0028] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0029] As used in this specification, the term "about" typically means + / -5% of the value, more typically + / -4% of the value, more typically + / -3% of the value, more typically + / -2% of the value, even more typically + / -1% of the value, and even more typically + / -0.5% of the value.

[0030] In this specification, certain embodiments may be disclosed in a range-bound format. It should be understood that this "range-bound" description is merely for convenience and brevity and should not be construed as a rigid limitation on the disclosed range. Therefore, the description of a range should be considered as having specifically disclosed all possible subranges and the individual numerical values ​​within those ranges. For example, a description of the range 1-6 should be considered as having specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., and the individual numbers within those ranges, such as 1, 2, 3, 4, 5, and 6. This rule applies regardless of the breadth of the range.

[0031] Definition of the noun: HRV data (heart rate variability data) refers to a measure of the time variation between each heartbeat.

[0032] PPG data is a method for measuring human physiological data that uses an LED light source and detector to measure the attenuation of light reflected back from the surface of human skin by blood vessels and other tissues, thereby recording the pulsation state of blood vessels and measuring the pulse wave.

[0033] EDA data (electrodermal activity data) refers to changes in the resistance or conductance of the skin surface, which reflects the activity of the sympathetic nervous system.

[0034] Blood pressure refers to the lateral pressure exerted on the wall of a blood vessel per unit area as blood flows through it; it is the driving force propelling blood flow within the blood vessels.

[0035] Blood oxygen saturation refers to the oxygen content in the blood, usually expressed as a percentage.

[0036] Sleep quality data refers to data calculated based on indicators such as sleep duration, deep sleep duration, and whether sleep is continuous, and is used to quantitatively assess a user's sleep quality.

[0037] Example 1: Please see Figure 3 , Figure 4 This invention provides a stress assessment method based on active physical interaction, which may include: S101: Acquire passive physiological data, wherein the passive physiological data includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data; S102: Calculate the physiological load level based on the passive physiological data; S103: Update the physiological load baseline based on scenario data; the physiological load baseline refers to a dynamic threshold that changes with the user's scenario and is used to determine whether the stress level is abnormal.

[0038] In some embodiments, the physiological load baseline is a threshold that dynamically changes with the contextual intent label (which may correspond to high / low stress scenarios) and is used to determine whether the user's stress level is abnormal.

[0039] In some embodiments, a physiological workload level above the baseline physiological workload can be considered a high physiological workload level, and a physiological workload level below the baseline physiological workload can be considered a low physiological workload level.

[0040] In some embodiments, the degree of abnormality of the physiological workload level can be determined based on the difference between the physiological workload level and the physiological workload baseline, and the stress level can be assessed based on the degree of abnormality of the physiological workload level. For example, if the difference between the physiological workload level and the physiological workload baseline is large (e.g., the difference between the physiological workload level and the physiological workload baseline is greater than or equal to a preset difference), then the degree of abnormality of the physiological workload level can be considered large, that is, the user's stress level can be determined to be high; conversely, if the difference between the physiological workload level and the physiological workload baseline is small (e.g., the difference between the physiological workload level and the physiological workload baseline is less than a preset difference), then the degree of abnormality of the physiological workload level can be considered small, that is, the user's stress level can be determined to be low.

[0041] In some embodiments, S103 includes: S1031: Acquire the scenario data, which includes active physical data. The active physical data refers to the rotation data generated by the wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotational angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. S1032: Generate at least one scenario intent label based on the scenario data; S1033: Update the physiological load baseline based on the scenario intent label; S104: Determine the user's stress level based on the physiological load baseline and the physiological load level.

[0042] It should be understood that the technical approach adopted in patent application CN120766975A is to calculate the difference between physiological data, physical data and preset benchmarks to obtain the first index and the second index respectively, and then to perform weighted fusion of the two to determine the stress state (that is, to determine the stress state based on the difference between physiological data, physical data and the corresponding benchmarks; if the difference is large, the user is considered to be under great stress).

[0043] In contrast, this invention prioritizes the relative levels of passive physiological data, using the sequence characteristics of active physical data as a reference for stress identification. Specifically, this invention differs from the aforementioned patent applications in at least the following technical approaches: 1) The stress level determination mechanism is dynamically adjusted based on actively input physical data as a discriminant factor of the user's situation (such as adjusting the physiological load baseline) in combination with the user's situation; rather than using physical data as a direct indicator for stress level calculation, and also not using the user's historical physiological average for stress level determination. 2) Use the sequence characteristics of physical data over a period of time (such as the average rotation speed and rotation pattern over a period of time) to dynamically adjust the physiological load baseline, rather than using the level of physical data (such as the speed) as the evaluation criterion.

[0044] In summary, this invention provides a technical solution for scenario differentiation based on physical data and for adjusting the stress assessment mechanism according to the scenario. This scenario differentiation scheme based on physical data can improve the accuracy of stress assessment.

[0045] Specifically, this invention distinguishes scenarios based on physical data, rather than directly calculating stress levels based on the levels of physical data. This improves accuracy; that is, scenario distinction can further enhance stress recognition accuracy based on users' personalized usage habits. For example, users who frequently use the rotation function may relieve stress in high-pressure environments (such as facing important meetings), pass the time when feeling down (such as encountering negative events or receiving negative news), or even meditate during leisure time (e.g., the ring has a counting function, and rhythmic rotation can assist users in meditation). Furthermore, when users are exercising, they often do not rotate the ring or rotate it only to a limited extent. In these different scenarios, the user's safe physiological load range may fluctuate to some extent.

[0046] When bored or at leisure, the physiological load baseline can typically be referenced to the historical average. However, when experiencing low mood, the physiological load baseline can be appropriately lowered (especially for users with underlying medical conditions who are not suited to significant emotional fluctuations), effectively increasing the alert level to provide timely relief solutions and prevent further escalation of negative emotions. Conversely, during exercise, the physiological load baseline can be relatively increased to avoid excessive alerts that could interfere with normal activities.

[0047] For example, some users may only use such functions occasionally, such as quickly rotating to decompress under high pressure, but they may also rotate them in their daily activities due to boredom or other reasons. In this regard, the method of scene recognition through sequence features can also avoid the impact of some unconscious or accidental rotation on the accuracy or reliability of pressure assessment. The applicant found that in the actual application of smart wearable devices, different user groups have significant differences in the frequency of use and dependence on their interactive functions (such as rotation and pressing). The scheme proposed in this application, which determines the scene based on physical data and then adjusts the pressure assessment mechanism based on the scene, can improve adaptability to different user groups.

[0048] In other words, this invention actually proposes different physical data (or active physical data) processing mechanisms and application methods for different user groups.

[0049] In other words, the technical approach adopted in this invention, which uses contextual data to generate contextual intent labels to explore users' real needs, can not only more accurately determine users' stress state, but is also particularly suitable for users who are accustomed to using the rotation function of wearable devices. That is, by using a large amount of active physical data generated by users based on their active interaction behavior, a model with contextual intent recognition capability is trained, thereby classifying and learning users' specific rotation behavior patterns, and thus exploring users' real stress state in specific situations.

[0050] In some embodiments, the wearable device may be a smart ring, a smart bracelet, or other wearable devices. The wearable device is configured with an outer ring and an inner ring, which are capable of rotating relative to each other.

[0051] In some embodiments, passive physiological data and active physical data can be obtained from sensors built into the wearable device.

[0052] For example, in some embodiments, magnetic elements are provided in the inner and outer rings respectively. When the inner and outer rings rotate relative to each other, the magnetic field formed between the magnetic elements changes accordingly. This change can be detected by a magnetic sensor, and the rotation angle of the inner / outer ring can be calculated based on the change in magnetic field, thereby calculating the number of rotations.

[0053] In some embodiments, a battery, a control circuit board, a wireless charger, and multiple sensors are disposed in the inner or outer ring. A sensory interaction module and an information acquisition module may be disposed within the inner ring. The sensory interaction module may include a Hall sensor, an inertial sensor, a touch sensor, etc.; the information acquisition module may include a heart rate sensor, a blood pressure sensor, a temperature sensor, an optical sensor, etc.

[0054] In some embodiments, the wearable device is also equipped with a temperature sensor to detect changes in the user's local body temperature and / or to collect the surface temperature of the skin in contact with the ring or the ambient temperature inside the ring.

[0055] In some embodiments, the wearable device is also equipped with a heart rate sensor to collect raw photoplethysmography (PPG) signals and calculate the real-time heart rate (bpm) using a built-in algorithm. It can perform single measurements and / or periodic / continuous measurements (such as sleep mode or exercise mode) according to user needs.

[0056] In some embodiments, the wearable device is also equipped with a blood oxygen sensor to acquire and calculate blood oxygen saturation and to estimate the user's systolic and diastolic blood pressure based on the characteristics of the blood oxygen signal.

[0057] In some embodiments, the wearable device is also equipped with an EDA sensor for collecting the user's skin conductivity.

[0058] In some embodiments, the wearable device is also equipped with a wearing status detection function, which can perform wearing status detection every certain period of time (e.g., 60 minutes) to ensure that the user wears the device correctly, thereby obtaining the user's passive physiological data and active physical data more accurately.

[0059] Among them, the wearing status detection can be determined by identifying whether the heart rate sensor can detect a valid heart rate signal and / or whether the temperature sensor reading is within the normal human body temperature range.

[0060] In some embodiments, sleep quality data may include the proportion of sleep stages (deep sleep / light sleep / REM), sleep continuity (number of awakenings / duration), and sleep efficiency (total time in bed to actual sleep time). Here, sleep quality data is not a raw measurement, but a composite score based on multi-dimensional features. The calculation process mainly includes converting each extracted sleep quality data point into a standardized score or grade, with higher grades representing greater sleep quality. For example, deep sleep score = (actual deep sleep duration / recommended deep sleep minutes) * 100; and sleep efficiency score = (total sleep time / total time in bed) * 100.

[0061] In some embodiments, sleep quality data can be calculated based on other existing technologies, such as the sleep quality calculation method in patent application CN106725327A, and the present invention does not limit it.

[0062] In some embodiments, HRV data, PPG data, EDA data, blood pressure data, blood oxygen saturation data, and sleep quality data acquired by wearable devices can be identified as passive physiological data.

[0063] Currently, traditional stress monitoring solutions using wearable devices generally rely on the analysis of users' heart rate variability (HRV data). HRV refers to the minute time differences between successive heartbeat cycles and is an important indicator for assessing the function of the autonomic nervous system. The autonomic nervous system consists of the antagonistic sympathetic and parasympathetic nervous systems. Normally, when the body is relaxed and at rest, parasympathetic activity dominates, the heart rate slows down, and HRV is higher, resulting in a flexible and varied heart rhythm. However, when the body faces challenges, feels tense, or is under stress, sympathetic activity increases, the heart rate accelerates, and the heart rhythm tends to be more uniform and rigid, leading to a decrease in HRV.

[0064] Therefore, traditional techniques generally adopt the stress assessment logic that "low HRV equals high stress, and high HRV equals low stress." However, in practical applications, the accuracy of its stress monitoring results is highly questionable by users, often showing a significant discrepancy with or even the complete opposite of users' subjective feelings. The applicant found that the root cause is that multiple factors can lead to a decrease in HRV, such as work anxiety, physical exercise, or caffeine intake. A decrease in HRV merely reflects the activation of the sympathetic nervous system and a high level of physiological load, but it cannot distinguish the fundamental source of this high level of physiological load based on passive physiological data.

[0065] For example, elevated EDA data could be due to emotional anxiety or excessively high ambient temperature. Similarly, other types of passive physiological data could be caused by a variety of factors.

[0066] Therefore, this invention proposes a stress assessment method based on active physical interaction, that is, identifying at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data as passive physiological data, calculating the user's physiological load level based on the passive physiological data, exploring the real causes of physiological load, and then comprehensively assessing the stress level by combining contextual data.

[0067] In other words, by using active physical data to verify stress levels, we can distinguish whether the source of physiological load is a high-stress factor or a low-stress factor, providing a reliable basis for stress level assessment. At the same time, by comprehensively assessing stress levels from two dimensions, namely internal physiological state (i.e., passive physiological data) and external behavioral performance (i.e., active physical data), we can greatly improve the accuracy of stress assessment results.

[0068] In some embodiments, the physiological load level can be calculated by the physiological load level calculation model based on the passive physiological data. Alternatively, a mapping table between the physiological load level and the passive physiological data can be preset in advance. For example, when the passive physiological data is greater than or less than the preset physiological data, the physiological load level is a high load level.

[0069] For example, in some embodiments, one type of passive physiological data, such as HRV data or blood pressure data, can be selected to calculate the initial physiological load level. If the blood pressure data exceeds a preset blood pressure threshold (which can be set with reference to the blood pressure data of a healthy population), the physiological load level is considered relatively high. Alternatively, the difference between the blood pressure data and the preset blood pressure threshold can be directly used to characterize the physiological load level.

[0070] For example, in some embodiments, two or more passive physiological data points can be selected to jointly calculate the physiological workload level. For instance, a pre-trained physiological workload assessment model can be used to generate the physiological workload level based on at least two passive physiological data points.

[0071] The physiological load level model (also known as a stress prediction model) can be trained using a historical physiological database. This database includes multiple training samples, and each training sample includes the values ​​of at least two recorded passive physiological data points, as well as the user's physiological load level under those conditions (this level can be assessed by a medical professional). Correspondingly, the input to the physiological load level model is the value of the passive physiological data, and the output is the physiological load level. Preferably, this physiological load level model can employ a random forest model.

[0072] Of course, in other embodiments, other deep learning models such as multilayer perceptron (MLP) or convolutional neural network (CNN) can be selected according to the prediction requirements.

[0073] It should be noted that the model in this embodiment can be trained using an existing deep learning model or implemented using an existing model; this invention does not impose any restrictions on this. For example, a stress prediction model from an existing smart wearable device (such as a smart bracelet) can be used. For instance, patent application CN201810186482.2 discloses a psychological stress monitoring system and method based on wearable devices and Android terminals.

[0074] In some embodiments, the physiological load level calculation model may be the neural state assessment model in the patent application with publication number CN120324744A.

[0075] Using HRV (Human Ventricular Reduction) and EDA (Electronic Depression) data as passive physiological data, a two-dimensional assessment model based on these two data sets can be constructed. The time-domain indices SDNN (standard deviation) and LF / HF (low-frequency to high-frequency power ratio) of HRV are used as inputs. When SDNN is below 20% of the user's baseline and the LF / HF ratio is above 30% of the baseline, it indicates significantly enhanced sympathetic nerve activity, meaning the HRV data is abnormal. Simultaneously, if the skin conductance level (SCL) in the EDA signal remains consistently elevated, or the frequency of skin conductance response (SCR) exceeds 5 times per minute, it further confirms a high level of physiological load. Therefore, the model can obtain a physiological load score ranging from 0 to 1 based on these two types of passive physiological data: HRV and EDA.

[0076] It should be noted that the physiological load level only means that "the body is in a high / low load state", but the reason is unknown and needs to be further determined by combining scenario data.

[0077] In some embodiments, the wearable device is also equipped with a Hall sensor to detect active physical data such as the number of rotations, rotation speed, rotation angular acceleration, rotation direction, and rotation angle of the user rotating the ring.

[0078] The rotation speed can refer to the average speed at which the user rotates the wearable device per unit time, or it can refer to the maximum or minimum speed.

[0079] Among them, rotational angular acceleration is used to characterize how quickly the rotational angular velocity of wearable devices changes.

[0080] In some embodiments, changes in the signal from the Hall sensor can be monitored in real time to determine whether the wearable device is rotating.

[0081] It should be understood that if the rotation direction is opposite to the user's habitual rotation direction, it can reflect a sudden change in the user's stress state to some extent. Of course, whether or not a sudden change in stress state actually exists needs to be judged in conjunction with other types of data.

[0082] A single rotation with a rotation angle greater than a preset rotation angle (e.g., 360°) can be defined as one revolution. It should be understood that the number of revolutions can be accumulated.

[0083] In some embodiments, the rotation detection algorithm within the sensory interaction module can perform effective de-jittering and / or filtering. For example, rotations that only undergo a small angular change (e.g., 10°) over a certain period of time (e.g., 3 seconds) are identified as erroneous rotations and are not included in the rotation count.

[0084] In some embodiments, S104 further includes: The contextual intent label can be generated based on a pre-trained contextual intent recognition model; wherein the contextual intent recognition model is trained through a historical contextual dataset, the historical contextual dataset including: the active physical data generated by the subject wearing the wearable device, and the corresponding contextual information.

[0085] In some embodiments, contextual intent tags can refer to dynamically generated semantic identifiers that reflect the user’s current state. They infer and characterize the user’s potential psychological state or behavioral purpose at a specific moment by fusing or selectively adopting available information from active interaction behavior patterns (active contextual intent tags) and passive environment or physiological context (passive contextual intent tags).

[0086] In some embodiments, the contextual intent label may be one of several pre-set rotation modes.

[0087] For example, if the active physical data over a certain period of time is high-frequency, small-angle, irregular reciprocating rotation, then the scenario intention label can be generated as an anxiety-type rotation pattern; if the active physical data over a certain period of time is low-frequency, low-speed, rhythmic, unidirectional, or regular full-circle rotation, then the scenario intention label can be generated as a relaxation-type rotation pattern.

[0088] In some embodiments, contextual intent labels are semantic labels generated by interpreting user-initiated physical behavior data through a pre-trained contextual intent recognition model. They do not merely describe the action itself (such as "turning"), but reveal the underlying psychological state or behavioral intent (such as "anxious" or "relaxed").

[0089] In some embodiments, when training a contextual intent recognition model, in addition to a general historical dataset, the model can also support: 1. Transfer Learning: a base model is pre-trained using a historical dataset, and then fine-tuned on the usage data of each user to quickly adapt to that user's unique behavioral patterns. 2. Few-Shot Learning: for a small number of users, if there is little applicable historical data to refer to, the model can be quickly adjusted through direct user feedback (such as several "yes / no" confirmations).

[0090] It should be understood that the contextual intent label is generated by the contextual intent recognition model, which is trained on a dataset containing historical active physical data and its corresponding real-world contextual information. This model can map raw physical signals captured from wearable devices (such as rotation direction, speed, etc.) to specific stress contextual categories, thereby providing key contextual verification information for interpreting physiological stress levels.

[0091] In some embodiments, physiological load levels are used to characterize the body's current level of activity or stress, but the reasons behind physiological load levels can be varied (it could be due to emotional stress, such as work anxiety; or it could be due to external environmental stimuli, such as high ambient temperature). Stress levels, on the other hand, are a further conclusion drawn from physiological load levels using situational intent labels determined based on situational data, i.e., to what extent physiological load is caused by psychological stress, thereby generating stress assessment results that can distinguish the sources of stress. Preferably, in this embodiment, situational intent labels are generated using average rotation speed and rotation patterns.

[0092] It should be understood that average rotation speed is quantified data that wearable devices (such as smart rings) can directly collect through common Hall sensors and photoelectric encoders. It does not require additional input from the user and can be used directly. At the same time, it can avoid the cumulative error introduced in subsequent calculations, thereby reducing computing power consumption.

[0093] Furthermore, the mapping relationship between the user's rotation speed and stress state is particularly clear. For example, under high stress, the user will unconsciously speed up the rotation (such as rapidly turning a ring when anxious), and is not easily disturbed by accidental movements. In other words, the rotation speed and rotation pattern can stably reflect the stress attributes under specific circumstances.

[0094] Preferably, the rotation law refers to the change pattern reflected by at least one of the number of rotations, the rotation speed, the rotation angular acceleration, and the rotation direction.

[0095] For example, taking the change pattern reflected by rotation speed as an example, the way to obtain the rotation law based on the rotation speed can be: acquiring a raw rotation speed data sequence over a period of time (e.g., one hour), the raw rotation speed data sequence can be divided into several subsequences according to a preset time length (e.g., ten minutes); calculating the change pattern feature values ​​between each subsequence (e.g., volatility, standard deviation from the average speed, periodicity, etc.); determining the rotation law reflected by the rotation speed based on the change pattern feature values ​​(e.g., if the standard deviation of the rotation speed between subsequences is large and there is no obvious periodicity, then the rotation regularity can be considered low); furthermore, the situational intention label can be determined as anxiety type based on the rotation law (e.g., low regularity).

[0096] In some embodiments, a user may be considered anxious based on low rotation regularity. Furthermore, the mapping relationship between rotation regularity and situational intent labels can be determined based on the specific performance of different users under different stress levels; this is not limited here.

[0097] In some embodiments, if multiple rotational data (such as the number of rotations and the rotational speed) are acquired simultaneously, the data can be comprehensively determined by fusing and analyzing multiple dimensions such as the standard deviation of rotational speed and rotational angle, the switching frequency of rotational direction, and the fluctuation of angular acceleration.

[0098] In some embodiments, the statistical values ​​(such as average values) of the user's physiological load levels under different default scenarios (such as work, rest) over a period of time (such as the first few days or weeks after wearing the wearable device) can be used as the initial physiological load baseline.

[0099] In some embodiments, updating the physiological load baseline based on the contextual intent label may include: collecting historical physiological load level data of the user under different preset contexts (such as during physical exercise and during high-intensity work); generating physiological load baselines for different contexts based on different historical physiological load levels under different contexts; and calling different physiological load baselines based on different contexts in which the user is located.

[0100] For example, statistical data on the user's first historical physiological load level under low stress conditions (such as when all passive physiological data are within the normal range), and data on the user's second historical physiological load level under high stress conditions (such as when at least one passive physiological data is outside the normal range).

[0101] Correspondingly, if the user is under low stress, first historical physiological workload data can be used to generate a first physiological workload baseline; if the user is under high stress, second historical physiological workload data can be used to generate a second physiological workload baseline. The value or level of the first physiological workload baseline is higher than that of the second physiological workload baseline. That is, when the user is healthy, the first physiological workload baseline has a higher tolerance for fluctuations in the user's physiological workload level.

[0102] Alternatively, the physiological load level calculation model can infer the range of physiological load levels under this scenario from historical physiological load level data; the upper limit of the physiological load level range or a specific value can be defined as the physiological load baseline under this scenario.

[0103] The physiological load level calculation model can refer to an algorithmic model that processes users' passive physiological data and quantifies the output to characterize the users' physiological load level. For example, it can be trained based on a random forest model or a CNN (convolutional neural network).

[0104] The input data may include at least one of the following: HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data.

[0105] The output data can be physiological load levels. In some embodiments, physiological load levels can be continuous values ​​(e.g., 0-100 points) or levels (e.g., low, medium, and high).

[0106] In some embodiments, different physiological load baselines can be invoked according to different scenarios of the user to determine whether the user's physiological load level is normal under different scenarios.

[0107] It should be understood that the physiological load baseline is not simply calculated based on historical data, but rather a personalized threshold that can change according to the user's situation.

[0108] In other words, the dynamic update mechanism for physiological load baseline provided by this invention sets different stress judgment criteria for different scenarios, that is, it adapts different stress recognition sensitivities to users in different scenarios, rather than directly equating physiological load with psychological stress. This achieves a certain degree of in-depth identification of stress sources and greatly reduces misjudgments. For example, when a user is running, the physiological load baseline is set relatively high (that is, the user's physiological load level can be temporarily at a high level in this scenario). At this time, even if the user's physiological load level is high, it will be compared with the high baseline in the exercise scenario, and thus will not be judged as high stress. Conversely, if the user is not exercising, the physiological load baseline is set relatively low (that is, if the user's physiological load level is high, it is an abnormal situation). At this time, the physiological load level can be compared with the low physiological load baseline in the meeting scenario, and it may be judged as a high stress state.

[0109] In some embodiments, the dynamic updating mechanism of the physiological workload baseline not only means that the physiological workload baseline can have different baseline values ​​under different scenarios, but also that it can be updated based on the user's performance in the same or similar scenarios over a long period of time. For example, for an employee who gradually goes from listening to speaking at a meeting to frequently speaking, as they become more familiar with the meeting environment, this psychological habit and adaptation will directly manifest as a significant downward trend in physiological workload levels during meeting scenarios. Therefore, the physiological workload level calculation model can identify this long-term, slow trend and update the physiological workload baseline (e.g., by lowering the physiological workload baseline). Simultaneously, if a physiological workload significantly higher than the baseline suddenly appears in a meeting that the user is already familiar with, it can be inferred that the user may be experiencing an abnormal stressful event (e.g., being suddenly questioned about an unprepared question).

[0110] In some embodiments, determining the user's stress level based on a combination of the physiological load baseline and the physiological load level may include: calculating the difference between the physiological load level and the physiological load baseline; if the result is greater than or equal to zero, it indicates that the user's physiological load level exceeds the physiological load baseline (or expected value) for that scenario, and the stress level can be determined to be high stress; furthermore, the greater the difference, the greater the stress; if the result is close to zero, the user can be considered to be in a normal state; if the result is less than zero, the user can be considered to be in a relaxed or low-arousal state, i.e., the stress level is low stress.

[0111] In some embodiments, S1032 may further include: Generate proactive contextual intent tags based on the proactive physical data; Passive contextual intent tags are generated based on the current time, geographical location, heart rate data, and the aforementioned sleep quality data; The context intent tag is generated based on the active context intent tag and the passive context intent tag.

[0112] In some embodiments, contextual intent recognition tags can be generated directly based on active physical data (including at least one of the following: number of rotations, rotation speed, rotation angular acceleration, and rotation direction, and the corresponding rotation pattern), current time, geographical location, heart rate data, and sleep quality data, using a contextual intent tagging recognition model.

[0113] Among them, the contextual intent label recognition model can be an algorithmic model used to analyze user contextual data, identify and output contextual intent labels that reflect the stress attributes of the user's current context. Its core function is to assist in updating the physiological load baseline and verifying the source of stress.

[0114] In some embodiments, the data input to the context intent label recognition model may include active physical data and / or scene data. The context intent label recognition model can output context intent labels that characterize the context in which the user is situated.

[0115] In some embodiments, the contextual intent label recognition model can be trained using a CNN or random forest model.

[0116] In some embodiments, the present invention preferably generates passive scenario intent tags based on the current time, geographical location, heart rate data, and the sleep quality data. That is, by integrating these four types of data, it is possible to more intelligently infer the user's stress level and underlying causes from multiple factors. For example, if a user's heart rate is high and lasts for a long time in the office on a weekday morning (current time) (geographical location) (heart rate data), but he only slept for 4 hours last night and only had 30 minutes of deep sleep (sleep quality data), then the increased heart rate is more likely due to physiological load and psychological irritability caused by insufficient sleep, rather than simply psychological stress, thus making a more accurate, comprehensive, and realistic judgment that fits the user's actual situation.

[0117] Different users exhibit circadian rhythms at different times of the day (e.g., a user is often alert in the morning and sleepy in the afternoon). The emotional significance of an elevated heart rate at 10 AM differs from that at 2 AM.

[0118] Geographic location is one of the most direct manifestations of different scenarios. Being in the office is typically associated with work stress or focus, while being at home is usually associated with rest and relaxation. Therefore, geographic location provides a crucial environmental reference for interpreting a user's stress level.

[0119] Among them, heart rate data is the most direct indicator of physiological arousal (whether it is positive excitement or negative anxiety).

[0120] Sleep is fundamental to physical and mental recovery. The quality of sleep the previous night directly determines a user's baseline ability to cope with stress that day. For example, a sleep-deprived user is more emotionally vulnerable, has a more volatile heart rate, and a significantly reduced tolerance for the same level of stress.

[0121] It should be understood that generating passive scenario intent labels based on current time, geographical location, heart rate data, and the aforementioned sleep quality data can more objectively assess stress levels and provide a more scenario-adaptive reference for the dynamic updating of physiological load baselines.

[0122] In some embodiments, contextual intent tags can be generated jointly based on active contextual intent tags and passive contextual intent tags and their corresponding weights. The active contextual intent tag has a greater weight than the passive contextual intent tag.

[0123] Alternatively, when active context labels conflict with passive context labels (e.g., the active context label is "anxious rotational pattern" and the passive label is "physical exercise"), the higher weight of the active context intention label allows the model to prioritize the active context intention label and generate the final high-stress context intention label.

[0124] It should be understood that the weight of active contextual intent labels is higher than that of passive contextual intent labels. This can better respect the subjective state conveyed by the user's spontaneous active rotation behavior, weaken the misunderstanding that the objective environment may bring, and improve the stability of the model's output stress level assessment results.

[0125] In other words, this embodiment combines user-specific performance (such as heart rate data and sleep quality data) with external objective conditions (such as current time and geographical location) to make a certain degree of reliable prediction of the user's current emotions, thereby serving as an auxiliary assessment condition for the level of physiological load.

[0126] For example, a passive contextual intent label might indicate that the user is at home (geographical location) and it's Saturday night (time), but an active contextual intent label might identify an anxious rotational pattern. Instead of immediately assuming work stress, the label might infer, based on a combination of active and passive contextual intent labels, that the user is likely dealing with family conflicts or contemplating difficult personal matters. This cross-validation mechanism improves the accuracy of the judgment and prevents misjudgments.

[0127] For example, a passive contextual intent label indicates that the user is in an important meeting (verified by calendar event and geolocation) and that the user did not get enough sleep the previous day. Simultaneously, an active contextual intent label shows an anxiety-driven rotation pattern. In this case, not only can a high-stress state be confirmed, but the source of stress can also be explained from multiple perspectives (i.e., not only work pressure, but also the user's poor condition due to sleep deprivation), potentially leading to more feasible stress-relief solutions (such as reminding the user to rest early that evening).

[0128] In some embodiments, if the user does not frequently rotate the device, it is insufficient to generate a corresponding active contextual intent tag based on active physical data. In this case, a variety of passive contextual intent tags (time, location, calendar, heart rate) can be relied upon to generate a relatively reliable contextual intent tag.

[0129] In some embodiments, if the passive contextual intent label is normal (location at home, time at leisure, good sleep), but the active contextual intent label continues to display an anxious rotation pattern, it may indicate that the user has latent anxiety or habitual tension. In this case, a more effective long-term stress-relief solution can be provided. Alternatively, if the user does not adopt this approach, the indicators corresponding to the anxious rotation pattern, such as the number of rotations, rotation speed, rotation angular acceleration, rotation direction, and rotation regularity, can be updated to better adapt to changes in the user's behavior.

[0130] In some embodiments, the physiological load baseline corresponding to the passive context intention label can be fine-tuned based on data that may affect the user's emotions, such as calendar event information (e.g., quarterly work reports), environmental data (e.g., ambient noise detected by the phone's microphone), weather data (e.g., whether it is raining or whether it is extreme weather), or behavioral pattern data (e.g., phone usage data reflecting that the user frequently switches applications or stays on a certain interface for a long time).

[0131] In some embodiments, by fusing active and passive contextual intent tags to generate contextual intent tags for the current context, the system can see both the user's "external situation" (passive) and hear the user's "behavioral intent" (active), which greatly improves the contextual adaptability of stress level assessment and helps to provide accurate, personalized and timely effective intervention for users.

[0132] It should be understood that passive contextual intent tags originate from continuous, passive observation of the user's state and environment, representing the user's situation and actual physical condition—objective background information. Active contextual intent tags, on the other hand, originate from the user's conscious, proactive behavior (such as rotating the device), representing the behavioral intent the user wants to convey by rotating the wearable device—a signal actively transmitted by the user to the device, and thus highly subjective. This invention combines passive and active contextual intent tags, with the core objective of improving the accuracy and contextual adaptability of stress level assessment through cross-validation.

[0133] In some embodiments, see Figure 5 This invention provides a stress assessment system based on active physical interaction, comprising: A passive physiological data acquisition module is used to acquire passive physiological data, which includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data. A physiological load level calculation module is used to calculate the physiological load level based on the passive physiological data; A physiological load baseline update module is used to update the physiological load baseline based on scenario data; the physiological load baseline refers to a dynamic threshold that changes with the user's scenario and is used to determine whether the stress level is abnormal; wherein, the physiological load baseline update module further includes: The scenario data acquisition unit is used to acquire the scenario data, which includes active physical data, which refers to the rotation data generated by a wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotation angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. A scenario intent tag generation unit is used to generate at least one scenario intent tag based on the scenario data; A physiological workload baseline update unit is used to update the physiological workload baseline according to the scenario intent label; The stress level determination module is used to determine the user's stress level based on the physiological load baseline and the physiological load level.

[0134] It should be understood that the stress assessment system based on active physical interaction can be used to perform the steps described in any embodiment of the present invention.

[0135] Example 2: In other embodiments, please refer to Figure 1 The present invention also provides a stress assessment method based on active physical interaction, comprising: S101: Acquire passive physiological data, wherein the passive physiological data includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data; S102: Calculate the physiological load level based on the passive physiological data; S103: Acquire contextual data, which includes active physical data. The active physical data refers to rotational data generated by a wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotational speed, rotational angular acceleration, and rotational direction. Both the passive physiological data and the contextual data are acquired from the same user wearing the wearable device. S104: Generate at least one scenario intent tag based on the scenario data; wherein, this includes: determining a corresponding preset scenario intent tag based on active physical data; S105: Determine the pressure level of the physiological load level based on the scenario intent label and the physiological load level.

[0136] In some embodiments, determining the stress level of the physiological workload level based on the scenario intent label and the physiological workload level may include the following steps: The scenario intent labels include a first type of scenario intent label (e.g., anxiety-induced rotation mode) that represents a preset high-stress scenario and a second type of scenario intent label (e.g., relaxation-induced rotation mode) that represents a preset low-stress scenario. The stress source for determining the stress level is based on the situational intent label and the physiological load level using preset adjudication rules, including: Decision rule (1): If the physiological load level is greater than or equal to the first threshold, and the situational intent label is identified as the first type of situational intent label, the stress source is determined to be a high-stress factor; Decision rule (2): If the physiological load level is greater than or equal to the first threshold, and the situational intent label is identified as the second type of situational intent label, the stress source is determined to be a low-stress factor; The pressure level is verified based on the pressure source.

[0137] In some embodiments, stress levels can be quantified into different grades, such as 1-10, where grades 1-5 represent low stress levels and grades 6-10 represent high stress levels. If the stress level does not match the stress source, such as a high stress level but a low-stress factor, the stress level can be updated, for example, by lowering the stress level.

[0138] For example, when a user's physiological load level is greater than or equal to a first threshold, and the situational intent label is identified as a type of situational intent label under a preset high-stress situation, that is, when the user's stress source is a high-stress factor (e.g., the user is attending an important meeting), the stress level can be maintained or increased.

[0139] Alternatively, when a user's physiological load level is greater than or equal to the first threshold, and the contextual intent label is identified as a type II contextual intent label under a preset low-stress scenario, i.e., the user's stress source is a low-stress factor (e.g., the user is exercising), the stress level can be reduced.

[0140] In other words, this invention provides a stress level verification method based on active physical data. It identifies the specific situation of the user based on active physical data, verifies and updates the stress level, so that the stress assessment results are more consistent with the user's real experience (although the body is tired and the physiological load is high after exercise, the mood is pleasant). At the same time, it effectively reduces false alarms, thereby providing users with more targeted and effective intervention measures.

[0141] In some embodiments, the physiological load level may be a fraction (e.g., 0.8, 0.9) or a grade (e.g., 1-10).

[0142] For example, if the physiological load level is 0.85 (i.e., the physiological load level is high), the scenario intent label generated is INTENT_ANXIOUS_FIDGETING (or it can be called an anxiety-type rotation pattern). The scenario data shows that the user is sitting still, and the calendar event information shows a project report. At this time, high physiological load and a clear anxiety behavior pattern appear simultaneously, and the scenario matches a high-stress scenario. At this time, the stress assessment result can be output as "high psychological stress", with an explanation: "Your body is detected to be in a state of high tension, and your ring interaction pattern shows typical anxiety characteristics." For example, if the physiological load level is 0.9 (i.e., the physiological load level is extremely high), the context intention label is generated as INTENT_NEUTRAL (or it can be called a relaxation rotation mode). The context data shows that the user is running and the geographical location is in the gym. At this time, the physiological load is extremely high, but there is no anxious behavior pattern. The context data clearly points to physical exercise. Therefore, the stress assessment result can be output as "physiological load (high-intensity exercise)" rather than a stress event.

[0143] For example, if the physiological load level is 0.6 (i.e., moderate physiological load), and the scenario intent label is INTENT_MINDFUL_COUNTING (or what could be called a mindfulness-based rotational mode), the scenario data shows that the user has activated the "breathing training" function in the app. At this point, the user's physiological load level has not yet fully returned to normal, but based on the user's behavior of activating breathing training, it can be inferred that the user is actively relaxing. Therefore, "Actively relaxing" can be output, and the user can be encouraged to continue. Subsequent monitoring of the changing trends in passive physiological data can verify the relaxation effect.

[0144] For example, if the physiological load level is 0.8, the contextual intent label is INTENT_NEUTRAL (or it could be called a relaxation-type rotation mode), and the contextual data shows that the time is late at night, the phone detects that the user is browsing a large number of posts about a certain social event on social media. At this time, the user's physiological load is high, but there is no anxious behavior. However, the contextual data suggests that there may be psychological stimulating factors. In this case, a more neutral result can be output, and options can be provided for the user to confirm, such as: "Your physiological load level is high. Are you paying attention to a certain social event?" The user's feedback can then be used as training data for the model.

[0145] In some embodiments, different high-stress scenario types and low-stress scenario types can be preset for different types of users.

[0146] For example, exercise can be identified as a high-stress situation or a low-stress situation; the specific classification can be determined based on the user's own stress levels.

[0147] It should be understood that this invention preferably presets personalized high / low stress scenarios for different types of users, avoiding the erroneous judgment of some users' excitement during exercise at the gym as stress, and also avoiding overlooking the implicit anxiety of other users in the home environment. That is to say, different users exhibit different individual stress responses in different scenarios; for example, some people experience high stress at work, while others experience the opposite. This invention identifies stress scenarios based on the relationship between the user's personalized scenario and stress level to output stress assessment results, thus more accurately reflecting the user's true subjective feelings.

[0148] In some embodiments, a contextual confirmation signal, such as “We have detected that you are under a lot of stress. Is this related to your current high-intensity mental activity, such as a meeting?”, can be issued to ask the user whether the contextual intent label is accurate.

[0149] In some embodiments, user feedback on situational confirmation signals will be recorded and used for: 1. Online fine-tuning, i.e., updating the user's situational intent recognition model in real time or periodically (e.g., strengthening the association between the situational intent label and the situational data when the user confirms the generated situational intent label), achieving personalized situational intent recognition based on the user's real feedback. 2. Error sample collection: adding misjudged cases (e.g., user denial of "anxiety-induced rotational pattern") to a specific error dataset for subsequent focused optimization and retraining of the model. 3. Stress situation library update: dynamically updating the user's personal "high / low stress situation" classification library based on user feedback.

[0150] In some embodiments, S105 further includes: Based on the pressure level, a corresponding decompression solution is generated, including: The pressure level is input into the decompression scheme generation model, which is used to generate the corresponding decompression scheme based on the pressure level. The decompression scheme includes guiding the user to adjust their behavior pattern through vibration of the wearable device and / or the display interface of other devices connected to the wearable device.

[0151] The decompression scheme generation model can be trained using a decompression scheme dataset, which includes pressure levels and corresponding decompression schemes.

[0152] In some embodiments, a decompression scheme can be generated by a decompression scheme generation model based on the input pressure level, or decompression schemes corresponding to different pressure levels can be preset in advance.

[0153] In some embodiments, the input data for the decompression solution generation model can be the user's stress level; the output data can be a specific decompression solution that the user can execute (such as the wearable device vibrating intermittently 3 times, each lasting 2 seconds, to guide the user to take deep breaths; or through a mobile APP connected to the wearable device, displaying the guiding text 'slowly rotate the ring for 1 minute while taking deep breaths' on the screen).

[0154] In some embodiments, the decompression scheme generation model can be trained based on a random forest model.

[0155] In some embodiments, stress levels and contextual intent tags can be combined to generate stress relief solutions. For example, when stress levels are high, if the contextual intent tag is anxiety-related and the user confirms they are in a meeting, the stress relief solution is preferably discreet, such as vibration-guided instruction only; if the contextual intent tag is relaxation-related and the user confirms they are after exercise, the stress relief solution could be a video of stretching exercises.

[0156] In some embodiments, the stress relief program can be adaptively adjusted based on the current time. For example, activities that refresh the mind are preferred during work hours, while programs that promote sleep and relaxation can be recommended at night.

[0157] In some embodiments, after detecting that a user has implemented a decompression program, the intervention effect of the program can be evaluated. For example, passive physiological data of the user can be collected again some time after implementing the program (e.g., 5-10 minutes later). The changes in physiological load levels before and after the decompression program intervention are compared. If the physiological load level decreases significantly, the decompression program is marked as "effective" for the user in this situation, reinforcing the association between the program and the situation; if the level does not change or increases, it is marked as "ineffective," and other alternative programs will be recommended when encountering similar situations in the future.

[0158] In some embodiments, see Figure 2 The accuracy of the physiological workload level calculation results (e.g., based on user evaluation) can be input into the physiological workload level calculation model, and / or the user's feedback on the situational intent label (e.g., acceptance or rejection) can be input into the situational intent recognition model, and / or the effect of the user's implementation of the decompression plan (e.g., changes in passive physiological data and active physical data before and after implementation) can be input into the decompression plan generation model to enhance the self-learning and optimization capabilities of the corresponding models. In some embodiments, the situational data further includes scene data; the scene data includes activity status data, and / or geographical location and / or calendar event information.

[0159] In some embodiments, scene data, as a component of contextual data, may further include at least one of current time, geographic location, heart rate data, and sleep quality data. Activity status data refers to the user's physical activity status, which can be collected by the inertial measurement unit (IMU) built into the wearable device, including sedentary activities, walking, exercise, and sleep.

[0160] Location data includes locations such as office, home, and gym. Calendar event information includes to-do items. Both location and calendar event information can be obtained from the terminal device connected to the wearable device.

[0161] In some embodiments, if the user does not respond to the context confirmation signal (e.g., the user is busy), the accuracy of the context intent label can be verified based on the context data.

[0162] For example, if the contextual intent label is "anxious rotation mode," and the geographical location in the contextual data is in the office, and / or the calendar event information is an important meeting, then the contextual data maps the user to be in a preset high-stress situation. The contextual data (high stress) and the stress level reflected by the contextual intent label (anxious rotation mode) are consistent, so the contextual intent label can be considered accurate.

[0163] In some embodiments, if the user inputs both active physical data and the wearable device acquires the user's scene data, the corresponding first contextual intent label can be determined based on the active physical data, and the corresponding second contextual intent label can be determined based on the scene data.

[0164] In some embodiments, the second contextual intent label can also be generated by a pre-trained contextual intent recognition model, which will not be elaborated here.

[0165] In some embodiments, the first scenario intent label and the second scenario intent label can be merged into a comprehensive scenario intent label and output; alternatively, they can be output separately, and then the stress level can be determined by combining the output scenario intent label and the physiological load level.

[0166] In some embodiments, stress levels can be determined based on passive physiological data, active physical data, and scenario data. For example, when the passive physiological data shows that the user is in a state of high physiological load, but the scenario intent label is not identified as "anxious rotation mode" and the acquired scenario data shows that the user is in a physical exercise scenario, the final stress assessment result is determined to be physiological load rather than high psychological stress.

[0167] It should be understood that through triple data cross-validation, the source of stress can be distinguished, minimizing misjudgments (such as misjudging physiological load caused by exercise as stress). Furthermore, stress relief solutions are no longer limited to a few fixed modes, but are dynamically adjusted based on the user's geographical location, calendar events, and other scenarios, ensuring the accuracy of stress perception and assessment by wearable devices while greatly improving the user experience.

[0168] In some embodiments, it also includes: Delete the passive physiological data and / or the active physical data that are less than or equal to a preset threshold.

[0169] It should be understood that deleting passive physiological data and / or active physical data that are less than or equal to a preset threshold can effectively distinguish noise from valid signals and reduce the false positive rate.

[0170] Specifically, for active physical data (i.e., rotational data), it can filter out unconscious, minute wrist movements (such as typing, waving, adjusting tension), ensuring that active physical data truly represents the user's conscious, intentional actions (such as specific rotational patterns during anxiety or relaxation), thereby greatly improving the accuracy of contextual intent labeling. For passive physiological data, it can eliminate subtle physiological signals caused by momentary poor device contact, making the calculated heart rate variability (HRV) and other physiological data more reliable, thus allowing the calculated physiological load levels to more closely reflect the user's actual state.

[0171] In some embodiments, a preset threshold can be increased when the user is detected to be exercising, thereby avoiding interference from motion artifacts (such as shaking after large movements) on physiological data and calculating a more accurate level of physiological load.

[0172] For example, if the accelerometer shows that the user is moving vigorously, even if there is a slight rotation, the model will prioritize judging it as an incidental action of the movement rather than an interaction with a clear intention, thereby improving robustness.

[0173] In some embodiments, different preset thresholds can be adaptively set according to the behavioral habits of different users, thereby adapting to the different physiological conditions (e.g., the physiological baseline values ​​of athletes and ordinary office workers are different) and behavioral differences (e.g., some people are naturally more prone to gestures), so that the stress assessment results are consistent and reliable under different users and diverse daily situations.

[0174] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0175] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0176] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A stress assessment method based on active physical interaction, characterized in that, include: S101: Acquire passive physiological data, wherein the passive physiological data includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data; S102: Calculate the physiological load level based on the passive physiological data; S103: Update the physiological workload baseline based on scenario data; The physiological load baseline refers to a dynamic threshold that changes with the user's situation and is used to determine whether the stress level is abnormal; wherein, S103 includes: S1031: Acquire the scenario data, which includes active physical data. The active physical data refers to the rotation data generated by the wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotational angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. S1032: Generate at least one scenario intent label based on the scenario data; S1033: Update the physiological load baseline based on the scenario intent label; S104: Determine the user's stress level based on the physiological load baseline and the physiological load level.

2. The stress assessment method based on active physical interaction according to claim 1, characterized in that, S1032 includes: A scenario confirmation signal is issued; the scenario confirmation signal is used to inquire whether the scenario intent label is accurate.

3. The stress assessment method based on active physical interaction according to claim 1, characterized in that, S1032 also includes: Generate proactive contextual intent tags based on the proactive physical data; Passive contextual intent tags are generated based on the current time, geographical location, heart rate data, and the aforementioned sleep quality data; The context intent tag is generated based on the active context intent tag and the passive context intent tag.

4. The stress assessment method based on active physical interaction according to claim 3, characterized in that, The rotation law refers to the change pattern reflected by at least one of the number of rotations, the rotation speed, the rotation angular acceleration, and the rotation direction; And / or, when generating the context intent tag, the weight of the active context intent tag is greater than that of the passive context intent tag.

5. The stress assessment method based on active physical interaction according to claim 1, characterized in that, Also includes: Based on the pressure level, a corresponding decompression solution is generated, including: The pressure level is input into the decompression scheme generation model, which is used to generate the corresponding decompression scheme based on the pressure level. The decompression scheme includes guiding the user to adjust their behavior pattern through vibration of the wearable device and / or the display interface of other devices connected to the wearable device.

6. The stress assessment method based on active physical interaction according to claim 1, characterized in that, The wearable device includes a smart ring with an outer ring and an inner ring, the outer ring and the inner ring being rotatable relative to each other.

7. The stress assessment method based on active physical interaction according to claim 1, characterized in that, The contextual data also includes scene data; the scene data includes activity status data, and / or geographic location and / or calendar event information.

8. The stress assessment method based on active physical interaction according to claim 1, characterized in that, Also includes: Delete the passive physiological data and / or the active physical data that are less than or equal to a preset threshold.

9. The stress assessment method based on active physical interaction according to claim 1, characterized in that, S1032 also includes: The contextual intent label is generated based on a pre-trained contextual intent recognition model; wherein the contextual intent recognition model is trained using a historical contextual dataset, which includes: the active physical data generated by the subject wearing the wearable device, and the corresponding contextual information.

10. A stress assessment system based on active physical interaction, characterized in that, include: A passive physiological data acquisition module is used to acquire passive physiological data, which includes at least one of HRV data, EDA data, PPG data, blood pressure data, blood oxygen saturation data, and sleep quality data. A physiological load level calculation module is used to calculate the physiological load level based on the passive physiological data; The physiological load baseline update module is used to update the physiological load baseline based on scenario data. The physiological load baseline refers to a dynamic threshold that changes with the user's situation and is used to determine whether the stress level is abnormal; wherein, the physiological load baseline update module further includes: The scenario data acquisition unit is used to acquire the scenario data, which includes active physical data, which refers to the rotation data generated by a wearable device that can be rotated by the user. The active physical data includes at least one of the following: the number of rotations of the wearable device, rotation speed, rotation angular acceleration, and rotation direction, as well as the rotation pattern. The passive physiological data and the scenario data are both acquired from the same user wearing the wearable device. A scenario intent tag generation unit is used to generate at least one scenario intent tag based on the scenario data; A physiological workload baseline update unit is used to update the physiological workload baseline according to the scenario intent label; The stress level determination module is used to determine the user's stress level based on the physiological load baseline and the physiological load level.

Images