A sleep and mood regulation control method and system based on specific spectral function

Abstract

The application provides a sleep and emotion regulation control method and system based on specific spectral function, and belongs to the technical field of medical devices, and specifically comprises the following steps: according to the regulation and control user data, and in combination with the consistency degree of the sleep characteristic signal of the regulation and control user in the light sleep stage, the regulation and control user is regulated and controlled, the determination of the suspension control method of the sleep and emotion regulation processing of the specific spectral function is adopted, the suspension control processing of the regulation and control user is carried out based on the suspension control method, according to the suspension control data of the regulation and control user, when it is determined that the construction processing of the personalized sleep staging model needs to be carried out, according to the suspension control data of the regulation and control user, the suspension matching condition of the regulation and control user is determined, based on the suspension matching condition and the stability condition of the sleep characteristic signal, the determination of the construction processing target of the personalized sleep staging model is carried out, and the reliability in the sleep and emotion regulation control process based on the specific spectral function is improved.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, and in particular relates to a sleep and mood regulation control device and method based on specific spectral functions. Background Technology

[0002] In modern society, sleep disorders and emotional problems (such as anxiety and depression) are highly commensurate. One of their common physiological bases is the disruption of circadian rhythms and neurotransmitter system disorders. To address these technical issues, CN202410140388.9, "A Near-Infrared Spectroscopy System for Monitoring Mental and Physical Stress and Sleep Quality," utilizes near-infrared spectral data to calculate and reflect the physiological characteristics of test subjects. It also combines classification algorithms to establish a sleep state recognition model and incorporates frequency domain variation technology to construct an analysis model. This allows for the analysis and evaluation of the test subjects' vital signs from multiple dimensions, and provides personalized massage experiences based on the evaluation results, offering users a more comfortable and personalized service experience. However, the above technical solution has the following technical problems:

[0003] In the process of controlling sleep and mood regulation based on specific spectral functions, clear division of sleep stages is crucial. Improper light stimulation during light sleep can affect the user's sleep quality. Therefore, how to build personalized models of light sleep stages for different users based on the termination delay during the control process, and determine the processing results, thereby improving the reliability of termination control processing, and providing more reliable experimental data for the control process, has become an urgent technical problem to be solved.

[0004] Therefore, there is an urgent need for a method and system for sleep and mood regulation control based on specific spectral functions. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted:

[0006] Specifically, this application provides a method for controlling sleep and mood regulation based on specific spectral functions, which includes:

[0007] S1 determines the sleep characteristic signals of the user's light sleep stage, and based on the identification and matching of the light sleep stage of the sleep characteristic signals of different dimensions, identifies the users who use sleep and emotion regulation processing with specific spectral functions, and uses them as the regulation users.

[0008] S3 determines the termination control method for sleep and emotion regulation processing using specific spectral functions for the regulated user based on the user's data and the consistency of sleep characteristic signals during the light sleep stage. The termination control process is then performed based on the termination control method. If, based on the user's termination control data, it is determined that a personalized sleep staging model needs to be constructed, the process proceeds to the next step.

[0009] S3 determines the termination matching status of the user based on the termination control data of the user, and determines the processing target for building a personalized sleep staging model based on the termination matching status and the stability of sleep characteristic signals.

[0010] The beneficial effects of this invention are as follows:

[0011] Based on the identification and matching of light sleep stages using sleep feature signals from different dimensions, users who employ sleep and emotion regulation processing with specific spectral functions are identified. This enables the screening of users with high reliability in light sleep stage identification and relatively stable sleep feature signals. It avoids the technical problem of high impact on users when using sleep and emotion regulation processing with specific spectral functions during light sleep due to high identification deviation in light sleep, thereby reducing the impact on users' sleep quality.

[0012] Based on the termination matching situation and the stability of sleep feature signals, the target for constructing a personalized sleep staging model is determined. This takes into account both the timeliness of light sleep stage identification and processing under the current model and the stability of sleep feature signals. By selecting construction targets with high timeliness and high stability, the accuracy and reliability of the suspension regulation processing during sleep and emotion regulation processing using specific spectral functions are further improved. This avoids the influence of external users' sleep features on the identification stability during the model's training iteration process, thus preventing the technical problem of poor identification reliability for users with high timeliness and high stability.

[0013] Furthermore, the sleep characteristic signal is determined based on the monitoring data from the sleep monitoring device during the user's sleep process.

[0014] Furthermore, the method for determining the user to be controlled is as follows:

[0015] Based on the identification and matching of light sleep stages in sleep feature signals of different dimensions, the sleep feature signals of light sleep stages identified in different sleep processes in history are determined and used as identification feature signals.

[0016] Sleep processes exhibiting distinctive signals are considered as sleep processes for identification.

[0017] Based on the identified sleep process and the identified characteristic signals during the sleep process, it is determined whether the user is a control user.

[0018] Furthermore, the method for determining the processing target in constructing the personalized sleep staging model is as follows:

[0019] Based on the user's discontinuation of matching, the duration of discontinuation after the user enters a light sleep period during sleep is determined and used as the duration of matching discontinuation during the sleep process.

[0020] Sleep processes with a matching termination duration within a preset termination duration range are considered as matching sleep processes. The combination data of the user's sleep processes are determined based on the stability of the sleep characteristic signals.

[0021] Based on the matched sleep process and sleep process combination data of the controlled user, it is determined whether the controlled user belongs to the target of the personalized sleep staging model construction.

[0022] It should be noted that if the user being regulated does not have a matching sleep process, then the user being regulated is determined not to be the target of the personalized sleep staging model construction.

[0023] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned sleep and mood regulation control method based on a specific spectral function when running the computer program.

[0024] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0026] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0027] Figure 1 This is a flowchart of a sleep and mood regulation control method based on specific spectral functions;

[0028] Figure 2 This is a flowchart of a method for regulating users;

[0029] Figure 3 This is a flowchart illustrating the method for determining the termination control method of sleep and mood regulation processing using specific spectral functions to regulate users;

[0030] Figure 4 This is a flowchart illustrating the process for determining the construction of personalized sleep staging models. Detailed Implementation

[0031] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0032] Example 1 Figure 1 As shown, this application provides a method for controlling sleep and mood regulation based on specific spectral functions, specifically including:

[0033] S1 determines the sleep characteristic signals of the user's light sleep stage, and based on the identification and matching of the light sleep stage of the sleep characteristic signals of different dimensions, identifies the users who use sleep and emotion regulation processing with specific spectral functions, and uses them as the regulation users.

[0034] S3 determines the termination control method for sleep and emotion regulation processing using specific spectral functions for the regulated user based on the user's data and the consistency of sleep characteristic signals during the light sleep stage. The termination control process is then performed based on this method. If, based on the user's termination control data and the user's specific spectral function adjustment deviation data, it is determined that a personalized sleep staging model needs to be constructed, the process proceeds to the next step.

[0035] S3 determines the update and control strategy for the sleep staging model of the user based on the identification and processing results of the sleep duration of the light sleep stage of the user.

[0036] Furthermore, the sleep characteristic signal is determined based on the monitoring data from the sleep monitoring device during the user's sleep process.

[0037] It is understood that the identification and matching results are determined based on the brain wave characteristics, heart rate characteristics, and breathing characteristics during different sleep processes. Specifically, the results are determined based on the sleep characteristic signals of the light sleep stage during different sleep processes.

[0038] This process aims to build a safe and effective pre-filter. Its core logic is: sleep intervention based on a specific spectrum will only be initiated for users whose system can reliably and consistently identify the single point in time when they enter a light sleep stage. This logic is based on a crucial premise: inaccurate light stimulation occurring during light sleep may disrupt sleep rhythms and actually reduce sleep quality. Therefore, the process comprehensively assesses the "technical feasibility" of the system intervening in a user's sleep by evaluating the success rate of successfully capturing the "entering light sleep" event in historical data and the reliability of the captured signal.

[0039] Specifically, such as Figure 2 As shown, the method for determining the user being controlled is as follows:

[0040] S11 determines the sleep characteristic signals of the light sleep stage in different sleep processes in history based on the identification and matching of light sleep stage of sleep characteristic signals of different dimensions, and uses them as identification characteristic signals.

[0041] The term "identification feature signal" is explained as follows: "Identification feature signal" refers to the specific physiological signal pattern (which may be identified by a combination of changes in EEG, heart rate, and respiratory characteristics) that marks the first entry into the light sleep stage (N1 or N2 stage) in a user's single night's monitoring data, based on a preset algorithm.

[0042] This serves as the logical starting point and data unit for the entire assessment. It transforms the complex overnight sleep stages into a clear, countable "event"—namely, "whether the turning point from falling asleep to entering light sleep was successfully captured." The significance of this definition lies in refining the timing of intervention triggers; the system only needs to respond to this one high-value point in time. If this signal cannot be identified throughout the entire sleep process, it means that the system has completely lost the "key" to initiating intervention.

[0043] Example: About 15 minutes after falling asleep last night, the system detected that user A's brain waves changed from alpha waves to theta waves, accompanied by a steady decrease in heart rate. This brain wave signal and heart rate characteristic signal were marked as the "identification characteristic signal" of this sleep.

[0044] S12 identifies sleep processes that exhibit characteristic signals as sleep processes.

[0045] S13 determines whether the user is a control user based on the identified sleep process and the identified feature signals during the identified sleep process.

[0046] It is understandable that if the user does not identify the sleep process, it is determined that the user cannot effectively identify the light sleep stage. Therefore, when using the specific spectrum function for sleep and mood regulation, once the user enters the light sleep stage without reliable identification, prolonged light interference may lead to poor sleep quality. Thus, it is determined that the user is not a regulation user, i.e., the specific spectrum function cannot be used for sleep and mood regulation.

[0047] The key to determining "sleep process identification" and preliminary screening is: "Sleep process identification" refers to the complete sleep record in which the "identification feature signal" was successfully captured and recorded during that sleep period.

[0048] This step elevates the evaluation dimension from "single event" to "historical record," assessing the system's long-term ability to capture the user's sleep inflection points. If the system fails to capture any signals of entering light sleep across all of a user's sleep records (i.e., the number of "identified sleep processes" is 0), it indicates that the system is completely unable to locate the intervention window. Initiating automatic spectral intervention in this case carries entirely uncontrollable risks. The significance of this step lies in performing initial safety interception, excluding users for whom the system is completely "blind."

[0049] Example: Over the past 7 nights, User A successfully captured characteristic signals in the early stages of falling asleep 5 nights (i.e., 5 "sleep detection processes"), while User B failed to capture any in the 7 nights (0 "sleep detection processes"). User B was directly identified as a "non-controlled user".

[0050] Additionally, it should be noted that if the user has a sleep recognition process in S131, the proportion of the sleep recognition process in the user's historical sleep processes is determined and used as the recognition proportion. It is then determined whether the recognition proportion is greater than a preset proportion threshold. If so, the user is determined to be a control user; otherwise, the process proceeds to the next step.

[0051] Preliminary assessment based on "recognition ratio":

[0052] Keyword Explanation: "Recognition Ratio" = (Number of recognized sleep processes / Total number of sleep processes). It measures the system's historical success rate in capturing the key event of a user "entering light sleep".

[0053] Setting a relatively high "preset threshold" (e.g., 80%) is to quickly identify highly compatible users. For these users, the system can reliably find intervention trigger points almost every night, indicating that their physiological patterns are highly compatible with the recognition algorithm, and the risk of misjudging the timing of intervention is extremely low. Directly identifying them as intervention users allows them to consistently benefit from precise interventions. This optimizes the experience for high-value users and improves system efficiency.

[0054] Example: User A successfully detected the signal of entering light sleep on 18 out of the past 20 nights, with a "recognition rate" of 90%, which is higher than the preset high threshold of 80%. The system directly identifies him as a "control user".

[0055] S132 determines whether the recognition ratio is within a preset ratio range (less than a preset ratio threshold). If yes, proceed to the next step; otherwise, determine that the user does not belong to the controlled user.

[0056] S133 determines whether there is a recognition sleep process with multiple recognition feature signals. If yes, proceed to the next step; otherwise, determine that the user does not belong to the control user.

[0057] Check the "recognition reliability" of users with a moderate success rate:

[0058] Keyword Explanation: "Reliable sleep process identification" specifically refers to records from a single sleep episode where the unique "identification feature signal" is identified through strong correlation and joint confirmation of feature signals from multiple independent dimensions such as EEG, heart rate, and respiration. This indicates that the identification has higher confidence and resistance to interference.

[0059] For users with a moderate recognition rate (e.g., 40%-80%), the system's ability to capture their sleep inflection points is inconsistent. This step aims to investigate the quality of the successfully captured events themselves. If a user is recognized on some nights, but all these recognitions rely on only a single or weak feature, it indicates that these successfully captured events may be highly random and susceptible to noise. Automatic intervention based on such low-confidence recognitions still carries significant risk. The significance of this step lies in performing secondary quality filtering to exclude users whose recognition results may be unreliable.

[0060] Example: User C was identified 12 out of the past 20 nights (60%), but further investigation revealed that these 12 identifications were all based solely on the single feature of slight changes in heart rate, without any support from EEG or respiratory features. Therefore, User C was classified as an "unregulated user".

[0061] S134 takes the sleep process with multiple identification feature signals as a reliable sleep process, and determines whether the user belongs to the control user based on the reliable sleep process and identification ratio of the user.

[0062] Specifically, based on the reliable identification of the user's sleep process and the identification ratio, it is determined whether the user belongs to the controlled user category, including:

[0063] The reliable identification ratio is determined based on the proportion of reliable sleep identification processes in the total sleep identification process of the user.

[0064] The average identification ratio is determined based on the average of the reliable identification ratio and the identification ratio, and the user is used to determine whether the user belongs to the controlled user group.

[0065] It is understood that when the average recognition ratio is greater than the preset recognition ratio threshold, the user is determined to be a controlled user.

[0066] The final review will be based on a comprehensive calculation of the "average recognition rate".

[0067] Keyword Explanation: "Average Recognition Rate" = (Recognition Rate + Reliable Recognition Rate) / 2. This is a composite metric that combines "quantity" (capture success rate) and "quality" (high-confidence capture rate) to more comprehensively evaluate the overall suitability for users with a moderate success rate.

[0068] For users with existing successful records, and some of those records being of high quality, careful consideration is required. The introduction of the "average recognition ratio" allows users with a decent success rate and high confidence when a record is captured to pass the evaluation. It balances the scenarios of "frequently capturing but occasionally misjudging" and "infrequently capturing but accurately capturing when successful." By calculating this average and comparing it with a suitable "preset recognition ratio threshold," a more refined and fair final review decision is achieved, ensuring that potentially valuable users are not unfairly penalized while strictly controlling risk.

[0069] Example: User D was identified 10 times in the past 20 nights (50% identification rate), of which 8 nights received strong confirmation from multiple physiological dimensions (80% reliable identification rate). Their average identification rate is (50% + 80%) / 2 = 65%. If the preset comprehensive threshold is 60%, then User D is ultimately determined to be a "manipulated user".

[0070] Furthermore, the controlled user data includes the number of controlled users.

[0071] This process aims to dynamically determine the safest and most suitable termination control strategy for the group identified as "regulatory users." Its core logic is to assess the risks and necessity of the system's "instant termination based on single-dimensional characteristics" safety mechanism, based on the size of the regulated user group and the homogeneity (consistency) of their sleep patterns, and select control methods of varying strictness accordingly. Its fundamental goal is to optimize the continuity of intervention and user experience by avoiding frequent and unnecessary interruptions to effective intervention due to an overly sensitive (conservative) termination strategy, while ensuring safety.

[0072] Specifically, such as Figure 3 As shown, the method for determining the termination control method of sleep and mood regulation processing using specific spectral functions for the user is as follows:

[0073] S21 determines the number of users to be controlled based on the aforementioned user data;

[0074] Determining the number of users to be controlled: The key term "number of users to be controlled" refers to the total number of users in the overall user pool who are ultimately determined to be suitable for spectral intervention through the aforementioned screening process.

[0075] This is the macro-level starting point for the assessment. User numbers are a fundamental scale indicator. If the proportion of users targeted is extremely small, it indicates that the current system algorithm or product model only suits a very small number of people. In this case, the effectiveness of intervention on a specific spectrum is difficult to determine effectively. In such situations, any intervention should adopt the most conservative strategy because the sample size is small, individual differences may not be fully understood, and potential unknown risks are high. The significance of this step lies in establishing the first level of risk assessment from the perspective of group size.

[0076] Example: In a user pool of 10,000 people, only 50 people were ultimately identified as control users.

[0077] S22 Based on the consistency of the sleep characteristic signals of the user in the light sleep stage, sleep processes with a similarity greater than a preset similarity coefficient threshold are grouped into the same sleep process combination.

[0078] The term "sleep process combination" refers to the clustering of "identification feature signals" (i.e., multi-dimensional feature patterns such as EEG, heart rate, and respiration during light sleep) extracted from all historical "identified sleep processes" of all users, based on the similarity of their waveforms, intensity, and timing. Sleep processes with a similarity higher than a preset threshold are grouped into the same combination.

[0079] This step is crucial for assessing group homogeneity. It goes beyond single-user analysis, aiming to discover whether common, stable physiological patterns exist within the user group. If a few "combinations" containing numerous sleep processes can be formed, it indicates that the group's sleep patterns are convergent, resulting in high predictability and generalization ability for system identification and intervention. Conversely, if there are numerous and dispersed combinations, it indicates diverse group patterns, and the system faces greater complexity and uncertainty. The significance of this step lies in quantitatively assessing the universality of intervention strategies.

[0080] Example: The 50 control users mentioned above provided a total of 5,000 "sleep process identifications". Through cluster analysis, these 5,000 processes mainly formed two major "sleep process combinations" (for example, combination A contains 3,000 processes; combination B contains 1,500 processes), while the remaining 500 processes were scattered in many small combinations or could not be classified.

[0081] S23 determines the controlled user based on the number of controlled users and the combined sleep process data of the controlled users, and adopts a sleep and emotion regulation processing termination control method with specific spectral functions.

[0082] Specifically, the sleep process combination refers to grouping sleep processes whose similarity of sleep feature signals in all dimensions is greater than a preset similarity coefficient threshold into the same sleep process combination.

[0083] It should be noted that, based on the number of users subject to regulation and the combined sleep process data of these users, the regulation users are determined, and a method for terminating sleep and mood regulation processing using a specific spectral function is employed, which specifically includes:

[0084] S231 Based on the number of users to be regulated, determine the proportion of the number of users to be regulated among the users, and determine whether the proportion is less than a preset proportion threshold. If so, determine to adopt a preset control method, determine the users to be regulated, and adopt a sleep and mood regulation processing termination control method with specific spectrum function. If not, proceed to the next step.

[0085] The primary decision-making process based on "quantity percentage" is explained as follows: "Quantity percentage" = (Number of users subject to regulation / Total number of users). It measures the prevalence of users who perform validation processing on the results of regulation of a specific spectrum within the overall user base.

[0086] Setting a low "preset percentage threshold" (e.g., 1%) is to identify "niche fit" scenarios. As shown in the example, 50 / 10000 = 0.5%, far below 1%. This confirms the initial judgment in step S21: this is a very small group. In such a "niche" scenario, frequent intervention and cessation of specific spectra will inevitably lead to experimental results that fail to meet requirements. Therefore, the most stringent "preset control method" (explained in detail below) is directly adopted to improve the reliability of the verification process for intervention and cessation of specific spectra.

[0087] Example: If the percentage of users subject to regulation is 0.5% (less than the preset threshold of 1%), the system will directly decide to apply the "preset control method" to all users subject to regulation.

[0088] S232 determines whether the number of sleep process combinations of the regulated user is less than the preset process combination number threshold. If so, it determines to adopt the second preset control method, determines the regulated user, and adopts the sleep and emotion regulation processing termination control method with specific spectrum function. If not, it proceeds to the next step.

[0089] S233 obtains the number of sleep processes in different sleep process combinations, and uses the proportion of the number of sleep processes in the sleep process combination to all sleep processes as the matching factor of the sleep process combination. It determines whether there is a sleep process combination with a matching factor greater than a preset matching factor threshold. If so, it determines to adopt the second preset control method, determines the user to be regulated, and adopts the sleep and emotion regulation processing termination control method with a specific spectrum function. If not, it proceeds to the next step.

[0090] Intermediate decision-making based on "portfolio concentration," key terms explained:

[0091] "Number of process combinations": refers to the total number of "sleep process combinations" obtained after clustering.

[0092] "Matching Factor": This refers to the proportion of sleep processes included in a given "sleep process combination" to the total number of sleep processes for all regulated users. It measures the representativeness of the combination.

[0093] These two steps assess the concentration of group patterns from different perspectives.

[0094] S232 (Small number of combinations): If the total number of combinations is very small (e.g., less than 3), it indicates that the group pattern is highly convergent and tends to be consistent. In such a highly homogeneous group, the system's recognition behavior is highly predictable, and the risk of misjudgment due to individual differences is low. Therefore, a relatively lenient termination strategy ("second pre-set control method") can be adopted to reduce unnecessary intervention interruptions.

[0095] S233 (Dominant Combination Exists): Even if the total number of combinations is large, as long as there is a dominant combination with an extremely high "matching factor" (e.g., >50%), it means that an overwhelming mainstream pattern exists. The system will have high confidence in identifying this mainstream pattern. For users who match the mainstream pattern, a relatively lenient termination strategy can also be applied. The significance of these two steps is that when the data shows that the group has high consistency or a strong dominant pattern, the safety restrictions can be appropriately relaxed to improve the continuity of intervention and user experience.

[0096] Example: If the proportion of users subject to regulation is high (e.g., 30%), and their sleep processes are clustered into only two main combinations (few combinations), then the system decides to adopt the "second preset control method".

[0097] S234 determines whether the user being regulated has a sleep process combination with matching factors within a preset matching factor range. If yes, proceed to the next step; otherwise, use a preset control method to determine the user being regulated and use a sleep and emotion regulation processing termination control method with a specific spectral function.

[0098] S235 takes the sleep process combinations with matching factors within a preset matching factor range as available combinations, determines the user to be regulated based on the sum of the matching factors of the available combinations, and adopts a sleep and emotion regulation processing termination control method with specific spectral functions.

[0099] Final decision based on "combined strength aggregation":

[0100] Keyword explanation:

[0101] "Preset matching factor range": A moderate range of proportions (e.g., 10% ≤ matching factor < 50%), used to screen out stable pattern sub-combinations that are not dominant but have a considerable scale.

[0102] "Available Combinations": Combinations of sleep processes whose matching factors fall within this range.

[0103] "Sum of Matching Factors": Sum the matching factors of all "Available Combinations".

[0104] This is a refined decision-making process designed to handle more complex "multimodal distribution" scenarios. When there is no single dominant combination, but there are multiple stable subgroups of considerable size (i.e., multiple "available combinations"), step S233 cannot make a decision.

[0105] S234: First, check if such a medium-sized subgroup exists. If even such a group does not exist, it indicates that the user pattern is extremely dispersed (long-tail distribution), with the highest uncertainty, and we should revert to using the strictest "preset control method".

[0106] S235: If multiple medium-sized subgroups exist, calculate their total influence (the sum of matching factors). If this sum is large (e.g., greater than 60%), it indicates that although the patterns are diverse, the system has been able to clearly identify and cover the stable patterns of most users. The system has a high degree of confidence in these clearly identified patterns, and therefore a lenient strategy ("second pre-defined control method") can be adopted for this "known majority." This reflects the refined management concept of implementing strategies hierarchically based on the degree of certainty that can be covered in complex distributions.

[0107] Example: Five combinations are generated to regulate user sleep processes, with no single dominant combination (all matching factors < 50%). However, three combinations have matching factors between 15% and 25% (belonging to the "usable combinations"), and their sum of factors is 15% + 20% + 25% = 60%, which is greater than the preset factor threshold of 55%. Therefore, the system decides to adopt the "second preset control method" for the user group whose 60% patterns are clearly covered.

[0108] It is understood that when the sum of the matching factors of the available combinations is greater than the preset factor threshold, the second preset control method is used to determine the user being regulated and to terminate the sleep and emotion regulation processing using a specific spectral function. Otherwise, the preset control method is used to determine the user being regulated and to terminate the sleep and emotion regulation processing using a specific spectral function.

[0109] It should be noted that the preset control method is to stop the sleep and mood regulation processing using the specific spectral function if, within the most recent preset time period, the analysis results of any sleep characteristic signal determine that the person is in the light sleep stage.

[0110] Specifically, the second preset control method is to stop the sleep and mood regulation processing using specific spectral functions as long as the analysis results of any sleep characteristic signal determine that the sleep is in the light sleep stage.

[0111] Preset control method (strict mode):

[0112] Rule: If, within the most recent preset time (e.g., the previous minute), any signal from any physiological dimension such as EEG, heart rate, or respiration that is analyzed in real time indicates that the user is "not in a light sleep stage", then spectral modulation will be stopped immediately.

[0113] Design Logic and Significance: This is an extremely conservative "presumption of guilt" strategy. It sacrifices the continuity of intervention, minimizing frequent interruptions, resulting in fewer experimental results for specific spectral interventions, and thus making effective verification difficult.

[0114] Second preset control method (relaxed / standard mode): Rule: Spectral modulation will stop as long as the physiological dimension signals monitored in any dimension of real-time analysis are consistent and continuously indicate that the user is "in a light sleep stage".

[0115] Design Logic and Significance: This is a relatively quick strategy for balancing intervention and termination, yet the reliability of the termination timing is ensured through a timely adjustment mechanism. It strikes a balance between security and user experience, and is suitable for user groups whose systems are well understood, whose patterns are stable, or whose user groups can be clearly categorized.

[0116] Specifically, such as Figure 4 As shown, the process of constructing a personalized sleep staging model is required, specifically including:

[0117] This process aims to assess the performance bottlenecks of the current general-purpose model by analyzing the timeliness of control termination, and to intelligently decide whether to initiate the construction of a personalized sleep staging model. Its core logic is: "Failure to terminate the process in a timely manner" (i.e., a slow system response after entering light sleep) is a direct manifestation of insufficient model accuracy. If such events are frequent or severe, it indicates that the general-purpose model cannot meet the needs of precise control. Forcibly employing stricter termination strategies (such as a "second preset control method") to compensate would excessively sacrifice intervention time; therefore, it is necessary to fundamentally improve staging accuracy by constructing a personalized model to achieve a balance between safety and effectiveness.

[0118] S31 determines, based on the control data of the user, that the user has entered the light sleep stage in different sleep processes, but has failed to stop the control in time, and records these as unstopped processes.

[0119] The term "untimely termination process" refers to a situation where, during a user's sleep cycle, precise post-sleep analysis indicates the user has entered a light sleep stage, but the system, using a real-time termination control method, determines that the actual time the termination is triggered exceeds a preset allowable delay range (preset duration range). For example, if the preset requirement is to terminate within 2 minutes of entering light sleep, and the actual termination time is 2 minutes and 30 seconds, then the process is flagged.

[0120] This step quantifies and refines the definition of the problem. It focuses not only on "whether it is stopped," but also on "whether it is stopped quickly enough." Delayed stopping means that the user received additional light exposure that should have been avoided during the light sleep stage, potentially amplifying the negative effects. Identifying these processes is a precise measure of the model's real-time performance from the perspective of "control effect," providing crucial data for evaluating the model's response sensitivity.

[0121] Example: The preset "timely" range is within 2 minutes of entering light sleep. In a certain intervention, user A actually entered light sleep at time T0, but the system stopped the process at T0+2 minutes and 30 seconds. This process was marked as "not stopped in time".

[0122] S32 uses the uninterrupted process data of the user to determine the termination control deviation factor of the user;

[0123] S33 determines whether a personalized sleep staging model needs to be constructed based on the user's cessation control deviation factor.

[0124] Calculate and evaluate the "Control Abortion Bias Factor." The term "Control Abortion Bias Factor" is applied to an individual user and is calculated as: (Number of "Untimely Aborted Processes" for the user) / (Total Number of Sleep Processes with Control Aborted by the user). It measures the frequency of "control delay" events experienced by the user.

[0125] This factor probabilizes the occurrence of delayed events and is a key indicator of whether a general model is adequately adapted to a specific individual. A high bias factor means that the user frequently experiences sluggish system responses, indicating a persistent mismatch between their physiological patterns and the recognition logic of the general model. This is a core individual warning signal that triggers personalized processing.

[0126] It should be noted that the sleep process that was not stopped in time refers to a sleep process that has entered the light sleep stage and whose interruption time is not within the preset time range.

[0127] Understandably, if different users do not stop the process, the stop control strategy needs to be further adjusted. At this time, the stability of the model is not high. If all users are treated with the second preset control method for stop control, it will inevitably lead to a shorter duration of sleep and mood regulation using specific spectral functions. Therefore, it is necessary to construct a personalized sleep analysis model.

[0128] Scenario 1: Widespread Problem ("Different control users all experienced failure to terminate the process in a timely manner"), Logic and Significance: If all or the vast majority of control users have experienced at least one instance of "failure to terminate in a timely manner," this proves that the currently deployed general termination control strategy (likely a "pre-defined control method" relying on a general phased model) has a fundamental, system-level lack of sensitivity. Its identification algorithm may be too conservative or have weak feature generalization ability, resulting in a slow response to all users.

[0129] Simply switching the strategy to a stricter "second preset control method" (stopping immediately upon detecting any signal indicating light sleep) might shorten the delay, but misjudgments (identifying non-light sleep as light sleep) would drastically compress the intervention duration, compromising overall effectiveness. The fundamental solution lies in establishing a highly sensitive, personalized staging model for each user, improving identification speed and accuracy from the source.

[0130] Example: 80% of users who implemented the control measures had a record of "failure to stop in time." This indicates that slow response is a common problem in the general model.

[0131] Additionally, it is understandable that if different users do not have uninterrupted processes, the proportion of uninterrupted processes in the sleep process of the interrupted control is used as a basis to determine the interruption control deviation factor. It is then determined whether there are any users whose interruption control deviation factor is greater than the preset deviation control factor threshold. If so, the interruption control strategy needs to be further adjusted. At this time, the stability of the model is not high. If all users are subjected to the second preset control method for interruption control, it will inevitably lead to a shorter duration of sleep and emotion regulation using specific spectral functions. Therefore, it is necessary to construct a personalized sleep analysis model. If not, proceed to the next step.

[0132] Some users have serious problems ("There are users whose control deviation factor is greater than the preset threshold (e.g., 25%)"). The logic and significance are as follows: Even if the problem is not widespread, there are individual users whose deviation factor is extremely high (e.g., more than 25%), which means that for these users, the general model is seriously mismatched and control delay has become the norm.

[0133] Why are personalized models needed? To ensure the experience and safety of some users, while guaranteeing the reliability of interference identification for specific spectral data and preventing them from continuously being exposed to the risk of delayed intervention, it is necessary to build customized models for them. This is a targeted reinforcement of the most vulnerable links.

[0134] Example: User B failed to stop the intervention in 5 out of 16 interventions, with a bias factor of 31.25% (>25% threshold), so personalized modeling must be initiated.

[0135] Obtain the percentage of users whose control process has not been terminated, and determine whether the percentage of users whose control process has not been terminated is less than the preset threshold for the percentage of users whose control process has not been terminated. If so, it is determined that no personalized sleep analysis model needs to be constructed. If not, proceed to the next step.

[0136] Determine whether the average value of the cessation control deviation factor for different control users is greater than the preset deviation factor threshold. If so, it is determined that a personalized sleep analysis model needs to be constructed. If not, it is determined that a personalized sleep analysis model does not need to be constructed.

[0137] Group assessment of problem scope and severity:

[0138] When the problem is neither widespread nor extremely severe, a more refined group assessment is required:

[0139] Check the scope of impact ("Percentage of users whose processes were not terminated in a timely manner < Preset percentage threshold (e.g., 15%)"):

[0140] If only a small number of users (e.g., <15%) experience occasional delays, this is considered an acceptable systemic error or a marginal case. Resource-intensive personalized modeling can be postponed.

[0141] Check the group mean deviation level ("mean stop control deviation factor > preset deviation factor threshold (e.g., 8%)"):

[0142] Significance: If the proportion of affected users is not extremely low, and their average deviation factor is significant (e.g., >8%), it indicates that the latency problem is quite severe and prevalent among the affected group. More importantly, the fact that "most users entered the light sleep stage" means that the core intervention scenario (light sleep) occurs frequently, accompanied by the latency problem. This exposes the performance shortcomings of the general model in the most frequent and critical scenarios. Therefore, to ensure the quality of control in the core scenarios, personalized modeling must be initiated to improve the accuracy and timeliness of the model in light sleep identification.

[0143] Example: 20% of users failed to abort the process in a timely manner, and the average deviation factor for these users was 15% (>8% threshold). This indicates that control delay is a significant problem among users who frequently enter light sleep, and the system determines that personalized modeling is needed.

[0144] This process implements a "model upgrade decision engine based on control timeliness defect analysis," the core value of which is:

[0145] In-depth monitoring from results to performance: By monitoring the key performance indicator of "termination delay", the effectiveness and reliability of the sleep staging model in real-time control scenarios are directly evaluated, which goes beyond simple ex-post accuracy analysis.

[0146] Identifying bottlenecks and guiding the correct optimization direction: The process clearly reveals a key dilemma: compensating for model latency by tightening termination rules (such as switching to the "second preset control method") comes at the cost of sacrificing effective intervention time. Therefore, the optimization direction is correctly guided to the fundamental path of "improving the model's own staging accuracy and speed," namely, building a personalized model.

[0147] Achieving tiered response and resource optimization: Based on the problem's prevalence (all users), severity (high deviation among individual users), or group significance (high average deviation among some users), the system intelligently decides whether to initiate comprehensive personalized modeling, target only the affected users, or postpone the process. This enables precise allocation of R&D resources.

[0148] Forming a safe, effective, and evolving closed loop: This process serves as the effectiveness verification and evolution trigger point for all preceding steps (user screening, termination strategy selection). It ensures that the system does not stagnate at a "usable but crude" general model stage, but continuously identifies shortcomings and proactively evolves towards a more precise and safer personalized service stage, ultimately providing each user with "just the right" lighting intervention.

[0149] Specifically, the method for determining the processing target in constructing the personalized sleep staging model is as follows:

[0150] This process aims to accurately select the most suitable and prioritized individual users from the user group that has been identified as "requiring personalized modeling" as the targets for personalized model building. Its core logic is to prioritize users who have already demonstrated high "predictability" and "stability" under existing general models as modeling targets. This is because building models for these users has a higher success rate and better results, enabling highly reliable identification under the "aggressive control strategy" (i.e., the second pre-set control method), thereby verifying the effectiveness of the personalized path and setting a precedent.

[0151] S41 determines the duration of the interruption of the matching process after the user enters a light sleep period during sleep based on the interruption of the matching process, and uses it as the duration of the matching interruption during the sleep process.

[0152] The "matching termination duration" is calculated. For a specific sleep process, the "matching termination duration" refers to the time interval between the system's precise post-event determination of the "entry into light sleep" and the actual moment the system triggers termination. It quantifies the actual delay of a single control operation.

[0153] This is a fundamental indicator for assessing the accuracy of control. By calculating the delay of each intervention, the user's sleep process can be precisely categorized according to the control effect, providing data for the subsequent screening of "well-controlled" sleep patterns.

[0154] S42 selects sleep processes with a matching termination duration within a preset termination duration range as matching sleep processes, and determines the sleep process combination data of the regulated user based on the stability of the sleep characteristic signal.

[0155] Identify "Matching Sleep Processes" and "Combinations of Sleep Processes." "Matching Sleep Processes" refers to sleep processes whose "Matching Stop Duration" falls within the "Preset Stop Duration Range." For example, if the preset "Good" range is [0 seconds, 1 minute 10 seconds], then processes delayed within 1 minute 10 seconds fall into this category. This represents a successful case of "Timely Control."

[0156] "Sleep Process Grouping": This method clusters processes based on the stability of "sleep feature signals" (i.e., multi-dimensional physiological patterns during light sleep) extracted from all of a user's sleep processes, grouping processes with similar features into the same group. This represents the internal consistency of the user's own physiological patterns.

[0157] This step simultaneously analyzes the user's historical data from two dimensions: control results and physiological patterns. "Matching sleep processes" identifies scenarios in which the existing general model works well for the user, while "sleep process combinations" reflects the stability of the user's own physiological characteristics.

[0158] The intersection of these two factors represents the gold standard for finding ideal subjects for personalized modeling: users with stable patterns and current general models that can already partially cover them. Building personalized models for such users is essentially improving accuracy on a "good baseline," with low risk and an extremely high success rate.

[0159] S43 determines whether the controlled user belongs to the target of the personalized sleep staging model construction based on the matched sleep process and sleep process combination data of the controlled user.

[0160] It should be noted that if the user being regulated does not have a matching sleep process, then the user being regulated is determined not to be the target of the personalized sleep staging model construction.

[0161] Furthermore, in step S431, if the user being regulated has a matching sleep process, it is determined whether the number of sleep process combinations of the user being regulated is less than a preset threshold for the number of process combinations. If yes, proceed to the next step; otherwise, it is determined that the user being regulated does not belong to the target of the personalized sleep staging model construction.

[0162] To filter users with "stable physiological patterns", the logic is to determine whether the number of "sleep process combinations" of a user is less than a small threshold (e.g., 3).

[0163] A small number of combinations indicates that the user's sleep physiology is highly stable and self-similar. Regardless of the day they sleep, their brainwaves, heart rate, and breathing patterns when entering light sleep are roughly the same. This is an ideal prerequisite for building a high-precision personalized model, as it improves the reliability of personalized sleep staging results rather than presenting a chaotic and variable phenomenon.

[0164] Example: User A's sleep records from the past 50 times show only two main clusters after "sleep feature signals." This indicates that his pattern is very stable, so proceed to the next step.

[0165] S432 determines the proportion of the number of matched sleep processes of the controlled user, and judges whether the proportion of the number of matched sleep processes of the controlled user is greater than the preset matching proportion threshold. If yes, it is determined that the controlled user belongs to the construction and processing target of the personalized sleep staging model. If no, proceed to the next step.

[0166] The logic for filtering users with "high success rate of control" is as follows: calculate the proportion of "matched sleep process" to the user's total sleep process (matching percentage) and determine whether it is greater than a high threshold (such as 70%).

[0167] A high matching rate means that, for this user, the existing general model can achieve relatively timely control in most cases. This indicates that: first, the general model has captured the user's main physiological characteristics; second, the user's physiological signals are of good quality and easy to capture. Building a personalized model based on this has a clear goal—to optimize "good control in most cases" to "precise control in all cases," which is an achievable and measurable improvement.

[0168] Example: User B's matching rate is 75% (>70% threshold). His case indicates high recognition reliability, which is the target for building a personalized sleep staging model.

[0169] S433 determines whether the user being regulated belongs to the target of the personalized sleep staging model construction based on the average duration of the matching interruption of the user in different sleep processes.

[0170] Furthermore, if the average duration of the matched interruption in different sleep processes of the user being regulated is within the preset interruption duration range, then the user being regulated is determined to be the target of the personalized sleep staging model construction.

[0171] The logic for filtering users with "controllable and stable delay levels" is as follows: For users whose matching rate does not reach a high threshold but whose physiological patterns are stable (passing S431), calculate the average "matching termination duration" of all their "matching sleep processes". Determine whether this average value still falls within the "preset termination duration range".

[0172] This step filters out a special group of users: while their success rate of "timely control" isn't extremely high (the matching rate is moderate), the latency of each successful control is very stable and short. This indicates that when the general model "captures" their signals, the response is rapid and consistent. These users may have a few "abnormal nights" (such as physical discomfort or significant interference) that lower their matching rate, but their core, stable physiological patterns are well understood by the general model and can be responded to quickly. Building personalized models for these users can strengthen their stable patterns and learn to identify abnormal patterns, thereby increasing the moderate success rate to a high success rate.

[0173] Example: User C's matching rate is 60% (below the 70% threshold), but on the nights he successfully controlled the system, the average termination delay was only 1 minute and 8 seconds (within the preset range of [0, 1 minute and 10 seconds]). This indicates that in his "standard mode," the system response is extremely fast and stable. He belongs to the modeling target of "good foundation and potential," that is, to build a personalized model for the user.

[0174] Through the above three-level screening, the final personalized sleep staging model is designed to process users who simultaneously possess the following characteristics:

[0175] The physiological patterns are inherently stable (few combinations of sleep processes) and have shown high controllability under the current general model (either a high success rate of matching or a short and stable control delay when successful).

[0176] The strategic significance of prioritizing these users over those with the most severe problems lies in:

[0177] Under strict control, the reliability and timeliness of the suspension control processing for some users can be guaranteed. At the same time, the sleep characteristics of other users are reduced in training the overall model, which may lead to bias in the model's identification of the aforementioned users, thus ensuring the overall reliability of the control.

[0178] In a second aspect, according to Embodiment 2, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned sleep and mood regulation control method based on a specific spectral function when running the computer program.

[0179] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0180] The foregoing has described specific embodiments of this specification. 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 the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0181] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for controlling sleep and mood regulation based on specific spectral functions, characterized in that, Specifically, it includes: The sleep characteristic signals of the user's light sleep stage are determined. Based on the identification and matching of the light sleep stage of the sleep characteristic signals of different dimensions, the users who adopt sleep and emotion regulation processing with specific spectral functions are identified and they are used as the regulation users. Based on the user data and the consistency of the user's sleep characteristic signals during the light sleep stage, the user is identified. A specific spectral function sleep and emotion regulation processing termination control method is used. Termination control processing of the user is performed based on the termination control method. When it is determined that a personalized sleep staging model needs to be constructed based on the user's termination control data, the process proceeds to the next step. Based on the cessation control data of the regulated user, the cessation matching status of the regulated user is determined. Based on the cessation matching status and the stability of sleep characteristic signals, the processing target for constructing a personalized sleep staging model is determined.

2. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The sleep characteristic signal is determined based on the monitoring data from the sleep monitoring device during the user's sleep process.

3. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The identification and matching results are determined based on the characteristics of brain waves, heart rate, and breathing during different sleep processes.

4. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The method for determining the controlled users is as follows: Based on the identification and matching of light sleep stages in sleep feature signals of different dimensions, the sleep feature signals of light sleep stages identified in different sleep processes in history are determined and used as identification feature signals. Sleep processes that exhibit identifiable characteristic signals are considered as sleep processes for identification. Based on the identified sleep process and the identified characteristic signals during the sleep process, it is determined whether the user is a control user.

5. The sleep and mood regulation control method based on specific spectral functions as described in claim 4, characterized in that, If the user does not identify the sleep process, it is determined that the user cannot effectively identify the light sleep stage, and therefore the user is not a controlled user.

6. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The user data for regulation includes the number of users subject to regulation.

7. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The consistency of sleep characteristic signals of the user during the light sleep stage is determined based on the deviation of sleep characteristic signals between different sleep processes.

8. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The method for determining the cessation control method of sleep and mood regulation processing using specific spectral functions for the user under regulation is as follows: Based on the aforementioned user data, the number of users subject to regulation is determined; Based on the consistency of sleep characteristic signals of the user during the light sleep stage, sleep processes with a similarity greater than a preset similarity coefficient threshold are grouped into the same sleep process combination. Based on the number of users subject to regulation and the combined sleep process data of the users subject to regulation, the users subject to regulation are identified, and a method for terminating sleep and mood regulation processing with specific spectral functions is adopted.

9. The sleep and mood regulation control method based on specific spectral functions as described in claim 1, characterized in that, The method for determining the processing target in constructing the personalized sleep staging model is as follows: Based on the user's discontinuation of matching, the duration of discontinuation after the user enters a light sleep period during sleep is determined and used as the duration of matching discontinuation during the sleep process. Sleep processes with a matching termination duration within a preset termination duration range are considered as matching sleep processes. The combination data of the user's sleep processes are determined based on the stability of the sleep characteristic signals. Based on the matched sleep process and sleep process combination data of the controlled user, it is determined whether the controlled user belongs to the target of the personalized sleep staging model construction.

10. A computer system, comprising: A memory and processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a sleep and mood regulation control method based on any one of claims 1-9.

Images