An insomnia cognitive management method and system with intelligent recommended sleep time

Abstract

This invention provides a cognitive management method and system for insomnia with intelligent suggestions for bedtime. The method includes: collecting and preprocessing multi-dimensional user data to obtain standardized input data; training a sleep analysis model based on historical datasets and continuously optimizing the model using user feedback data; using the sleep analysis model to perform inference based on real-time user data for the day, dynamically generating and outputting a personalized sleep plan including suggested bedtimes; obtaining actual sleep details and calculating sleep quality indicators; conducting continuous risk monitoring based on sleep quality indicators and triggering tiered intervention suggestions; and feeding the sleep quality indicators and user feedback data as training data into the model optimization step to achieve closed-loop optimization. This invention solves the problems of rigid suggestions, lack of personalization, and insufficient continuous optimization capabilities in existing technologies, achieving precise, dynamic, and intelligent sleep management.

Description

Technical Field

[0001] This invention relates to the field of sleep medical information technology, and in particular to a cognitive management method and system for insomnia that provides intelligent suggestions on sleep time. Background Technology

[0002] Insomnia is a significant problem affecting the physical and mental health of modern people. Currently, there are some applications or systems on the market to assist in sleep management, but most of them have relatively basic functions.

[0003] The first category consists of sleep data recording products. Their main function is limited to collecting and displaying basic data such as sleep duration, sleep onset time, and bedtime through wearable devices or manual input. These products lack in-depth analysis and interpretation of the data, cannot provide users with valuable improvement suggestions, and users find it difficult to make effective adjustments to their sleep behaviors based on this information.

[0004] The second category is general sleep advice products, which are usually based on a set of fixed rules (such as "adults should sleep for 8 hours") and push homogeneous sleep advice to all users, such as going to bed at a fixed time and avoiding caffeine. Because they do not take into account the individual differences of users, such as age, lifestyle habits, and daily activity levels, such advice is not targeted or effective enough and is difficult to adapt to the personalized needs of users.

[0005] The third category consists of products that use simple machine learning techniques. These products may be able to classify sleep quality (e.g., good, medium, poor), but their function stops at assessment. They do not form a closed loop of "assessment-suggestion-feedback-optimization". The system cannot generate dynamic and executable personalized sleep plans based on the assessment results, nor can it use subsequent user feedback data to continuously optimize the model or suggestion strategy.

[0006] In summary, existing technologies generally suffer from problems such as limited data dimensions, static and rigid suggestions, and a lack of personalized adaptation and continuous learning capabilities. As a result, users find it difficult to obtain accurate, dynamic sleep management solutions that can be adjusted according to their own changes in condition, leading to limited sleep improvement effects. Summary of the Invention

[0007] This invention aims to address the lack of intelligent analysis and personalized suggestion capabilities in existing insomnia cognitive systems, providing an insomnia cognitive management method and system with intelligent suggestions for sleep onset time. This invention constructs a professionally trained sleep analysis model, integrates multi-dimensional user data, and achieves intelligent analysis of user sleep quality and dynamic generation and optimization of personalized sleep plans. This reduces the professional knowledge bias of user self-adjustment, improves the targeting and effectiveness of sleep intervention, and enables continuous iteration of system performance.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a cognitive management method for insomnia with intelligent suggestions for sleep onset time is disclosed, comprising the following steps: S1. Collect multi-dimensional data from users, including several types of user basic information, daily behavior data, sleep data, and external treatment plan data; S2. Preprocess the collected multi-dimensional data to obtain standardized input data; S3. Train a sleep analysis model based on historical datasets and optimize the sleep analysis model using user feedback data. The historical datasets contain a large amount of users' sleep medical data and insomnia treatment case studies. S4. Based on the real-time user data in the standardized input data obtained in step S2, call the sleep analysis model to perform inference, dynamically generate and output personalized sleep plans. S5. Obtain the user's actual sleep details for the previous sleep cycle and calculate sleep quality indicators based on the actual sleep details; S6. Based on the sleep quality index calculated in step S5, continuous monitoring is carried out. When the monitoring data reaches the preset risk warning threshold, the corresponding graded intervention suggestions are triggered. S7. The sleep quality index obtained in step S5 and the user feedback data collected in step S6 are used as training data and fed back to step S3 to optimize the parameters of the sleep analysis model and achieve closed-loop optimization of the system.

[0009] In some implementations, in step S2, preprocessing includes data cleaning, format conversion, and outlier removal.

[0010] In some implementations, outlier removal specifically employs the Z-score algorithm for detection and removal, and normalizes multi-source data to unify the data format.

[0011] In some implementations, when training the sleep analysis model in step S3, the constructed model input feature vector includes at least the following four dimensions of associated data: User basic attribute dimensions, including age and body mass index; The daily behavior time series dimension includes time series data on exercise volume and dietary intake events; Historical sleep feedback dimensions include historical sleep efficiency and sleep latency; The external intervention tag dimension includes features of validated treatment protocols that match the user profile.

[0012] In some implementations, in step S7, the training data fed back to step S3 includes: Objective feedback data: Changes in sleep efficiency calculated in step S5; Subjective feedback data: User satisfaction or behavioral feedback indicators of the personalized sleep plan generated in step S4 or the graded intervention suggestions generated in step S6; The sleep analysis model uses both objective and subjective feedback data to construct a reward function for reinforcement learning in order to optimize the model parameters.

[0013] In some implementations, the sleep analysis model is built on the Transformer architecture, and continuous optimization is specifically achieved by setting a feedback data accumulation trigger mechanism, which triggers an incremental update of the model parameters every time a predetermined number of valid user feedback data are accumulated.

[0014] In some implementations, the predetermined quantity is 100.

[0015] In some implementations, step S4, invoking the sleep analysis model for inference, specifically includes: Compare and analyze real-time user data with pre-stored historical user behavior patterns; Identify deviations in daily behavior relative to historical patterns; Based on deviation characteristics and the current time, the recommended bedtime and type of sleep companion content are adjusted in real time through a sleep analysis model.

[0016] In some implementations, in step S4, the user's real-time data includes at least the amount of exercise, diet, and commuting time for the day, and the personalized sleep plan includes a suggested bedtime and its allowable error range, as well as recommended sleep companions. The response delay from data input to plan generation does not exceed 10 seconds.

[0017] In some implementations, a tolerance range of ±15 minutes is recommended for bedtime.

[0018] In some implementations, in step S5, actual sleep details include actual bedtime, actual sleep onset time, wake-up time, wake-up time, and nighttime wake-up time; sleep quality indicators include sleep efficiency, total sleep time, time spent in bed, and time spent before falling asleep.

[0019] In some implementations, in step S6, the preset risk warning threshold is a multi-level threshold, specifically including: If sleep efficiency falls below the first threshold for the first consecutive scheduled days, a level one warning is triggered and a medical advice is generated. When sleep efficiency falls below the second threshold for the second consecutive predetermined number of days, a level 2 warning is triggered and a medical device recommendation is generated. When sleep efficiency falls below the third threshold for the third consecutive predetermined number of days, a level 3 warning is triggered and a reminder to adjust sleep schedule is generated.

[0020] In some implementations, the first predetermined number of days is 3 days and the first threshold is 60%; the second predetermined number of days is 5 days and the second threshold is 70%; the third predetermined number of days is 7 days and the third threshold is 80%.

[0021] According to a second aspect of the present invention, an insomnia cognitive management system with intelligent sleep time suggestion is also disclosed, which is used to implement the above-mentioned insomnia cognitive management method with intelligent sleep time suggestion, the system comprising: The data collection module is used to collect multi-dimensional user data; The data preprocessing unit, connected to the data collection module, is used to clean, convert the format of, and remove outliers from the collected data. The model training module is used to build and train the sleep analysis model and continuously optimize the model parameters using feedback data. The user plan formulation module is connected to the data preprocessing unit and the sleep analysis model, and is used to generate personalized sleep plans based on real-time user data by calling the sleep analysis model. The sleep information statistics module is used to obtain actual sleep details supplemented by users and calculate sleep quality indicators; The data analysis module, connected to the sleep information statistics module, is used to monitor sleep quality indicators and generate graded intervention recommendations based on multi-level risk warning thresholds. Data storage unit is used to store user data, model parameters, historical schemes and feedback data; The outputs of the sleep information statistics module and the data analysis module are fed back to the model training module, forming a closed-loop optimization system.

[0022] The beneficial effects of this invention are: 1. Through an original multi-dimensional data fusion training method, the sleep analysis model can deeply understand the relationship between individual user differences and sleep status, solving the problems of single data dimensions and poor model generalization ability in existing technologies, and generating more personalized and adaptable suggestions.

[0023] 2. Through dynamic solution generation technology based on real-time data, it can respond to changes in the user's daily status and provide the best suggestions "for the day and at the moment". This is different from static and fixed suggestion modes, and improves the accuracy and timeliness of sleep management.

[0024] 3. By constructing a closed-loop optimization system that includes data feedback and model retraining, the system has the ability to learn continuously and improve itself as user data accumulates, thus achieving a sustainable improvement in the accuracy of suggestions.

[0025] 4. Through a multi-level risk warning mechanism, it can proactively identify the deteriorating trend of users' sleep health and provide step-by-step suggestions from self-adjustment to medical intervention, realizing the transformation from passive recording to proactive health management. Attached Figure Description

[0026] Figure 1 This is a framework diagram of the insomnia cognitive management system with intelligent suggestions for sleep time according to the present invention; Figure 2 yes Figure 1 The diagram shows a simplified operation flow of an insomnia cognitive management system with intelligent suggestions for sleep onset time. Figure 3 This is one of the flowcharts of the insomnia cognitive management method with intelligent suggestion of sleep time in this invention; Figure 4 This is the second flowchart of the insomnia cognitive management method with intelligent suggestions for sleep time according to the present invention; Figure 5 This is the third flowchart of the insomnia cognitive management method with intelligent suggestions for sleep time according to the present invention. Detailed Implementation

[0027] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0028] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0029] This invention belongs to the interdisciplinary field of medical informatics and artificial intelligence, specifically relating to a cognitive management method and system for insomnia with intelligent suggestions for sleep time.

[0030] like Figure 1-2 As shown, a cognitive management system for insomnia with intelligent sleep time suggestions is disclosed. This system mainly includes the following core modules, which work closely together to form a complete closed loop of "data collection - intelligent analysis - solution generation - effect feedback - model optimization": Data collection module: Function: As the "sensory" layer of the system, it is responsible for collecting comprehensive data that constitutes the user's sleep ecosystem from multiple channels and multiple modalities.

[0031] Implementation Details: This module integrates multiple data interfaces to collect multi-dimensional user data, including basic user information, daily behavior data, sleep data, and external treatment plan data. For basic user information (such as age, gender, height, weight, and occupation), a registration form is provided for one-time user input. For daily behavior data, it connects to smart bracelets and mobile health apps via Bluetooth, Wi-Fi, and other communication protocols to automatically synchronize time-series data such as steps, heart rate, and GPS location (to infer commuting). A manual input interface is also provided for users to record dietary details (such as caffeine and alcohol intake), subjective emotions, and stress levels. For sleep data, it connects to smart mattresses, non-contact sleep monitoring radar, and other devices via a dedicated API to acquire raw signals such as heart rate variability (HRV), respiratory waveforms, and body movement frequency. All data is sampled at a frequency of at least once per minute and includes a precise timestamp to ensure real-time data accuracy and continuity.

[0032] Data preprocessing unit (belonging to the data collection module): Function: As the "quality inspection and standardization" link of the system, it ensures that the data input to the downstream model is reliable in quality and uniform in format.

[0033] Implementation Details: The data preprocessing unit is connected to the data collection module and is used to clean, convert, and remove outliers from the collected data. This unit performs three levels of processing on the raw data stream: First, data cleaning is performed, using interpolation to fill in missing values ​​caused by short-term signal loss; next, outlier removal is performed using the Z-score algorithm to calculate the moving average and standard deviation for each feature dimension, filtering out data points exceeding the μ±3σ range as false alarms or physiological extremes; finally, critical data normalization is performed: numerical features are normalized to maximum and minimum, mapping all values ​​to the [0,1] interval; categorical features are one-hot encoded; and timestamps are converted into periodic features that are easy for the model to process (such as time of day, day of the week). The output is a well-structured feature vector with uniform dimensions, temporarily stored in a buffer.

[0034] Model training module and data storage unit: Function: Together, these two components constitute the system's "brain" and "memory" center. The data storage unit is the system's core database, employing a distributed architecture to persistently store all raw data, preprocessed features, historically generated schemes, user feedback, and parameters and hyperparameters of different versions of the sleep analysis model. The model training module is responsible for building, training, and iterating the sleep analysis model.

[0035] Implementation Details: The training process is divided into two phases. Phase 1 (Pre-training): A massive historical dataset (containing sleep medical records of over 100,000 users and 50,000 clinically validated treatment plans) is retrieved from the data storage unit to train an initial model based on the Transformer architecture using supervised learning. This model learns to predict sleep quality from complex, multi-dimensional features and generate a rudimentary form of intervention suggestions. Phase 2 (Online Reinforcement Learning Optimization): This is the core of the closed loop. Once the system is operational, the model training module continuously monitors the feedback data stream. It sets up a trigger mechanism; for example, every 100 valid feedback responses, an incremental learning iteration is automatically initiated. This module uses reinforcement learning algorithms (such as PPO) to construct a reward function from the feedback data (such as sleep efficiency improvement values ​​and user adoption of suggestions), automatically adjusting the model parameters so that the model can produce better outputs in the next decision. The optimized model parameters are immediately stored back in the data storage unit, completing the knowledge update.

[0036] User plan creation module: Function: As the "decision execution" layer of the system, it is responsible for generating and pushing highly personalized sleep plans at fixed times every day.

[0037] Implementation Details: This module connects to the data preprocessing unit and the sleep analysis model, and is automatically activated at a preset time (e.g., 6 PM). It retrieves the user's normalized real-time data (e.g., "exercise consumption: 850 kcal," "clock-out time: 6:15 PM") from the data preprocessing unit and combines it with the user's long-term profile retrieved from the data storage unit. Subsequently, it calls the current optimal version of the sleep analysis model for millisecond-level inference. The model not only outputs suggestions, but in a preferred embodiment, its internal attention mechanism can identify deviation features such as "today's activity level is significantly higher than the average," thereby dynamically adjusting the output. The final structured solution generated is as follows: {"recommended_bedtime": "23:30", "tolerance": "±15min", "content_type": ["audio", "article"], "content_id": ["relax_001", "sleep_science_005"]}.

[0038] This module ensures that the end-to-end response delay does not exceed 10 seconds and pushes the solution to the user through the application interface, SMS or email.

[0039] Sleep information statistics module and data analysis module: Function: Together, the two constitute the "Effect Evaluation and Health Early Warning" subsystem of the system.

[0040] Implementation details: The sleep information statistics module is used to obtain actual sleep details supplemented by users and calculate sleep quality indicators. For example, it provides a user-friendly morning feedback interface, where users input the specific times they actually went to bed, actually fell asleep, woke up during the night, and finally woke up. Based on this, the module automatically calculates key indicators such as sleep latency, total sleep time, sleep efficiency (a core indicator, calculated as: total sleep time / time in bed × 100%), and sleep fragmentation index, and presents them to users in the form of visual charts.

[0041] Data Analysis Module: This module connects to the sleep information statistics module to monitor sleep quality indicators and generate tiered intervention suggestions based on multi-level risk warning thresholds. Specifically, this module runs asynchronously in the background, continuously analyzing time-series data of all users' sleep efficiency and other indicators. It has a built-in multi-level risk warning logic (e.g., a level 1 warning for 3 consecutive days of efficiency less than 60%, a level 2 warning for 5 consecutive days of efficiency less than 70%, and a level 3 warning for 7 consecutive days of efficiency less than 80%). Once a threshold is triggered, it automatically generates intervention suggestions of the corresponding level. For example, a level 1 warning suggests "consult a sleep doctor as soon as possible," a level 2 warning suggests "consider trying a white noise sleep aid," and a level 3 warning pushes an article on "sleep adjustment techniques." Simultaneously, this module collects user feedback on these warnings (such as clicking "learn more" or "ignore").

[0042] like Figure 3-5 As shown, the present invention also discloses a cognitive management method for insomnia with intelligent suggestions for sleep time.

[0043] The following combination Figure 3-5 The flowchart shown illustrates in detail the specific implementation steps of the insomnia cognitive management method with intelligent sleep time suggestion of the present invention.

[0044] Step S1: Multi-dimensional user data collection: This step is performed by the data collection module, where the system automatically collects various types of data that constitute the user's sleep ecosystem from multiple sources. Specifically, this includes: User basic information, such as age, gender, height, weight, occupation and other static attributes, is entered and stored in the data storage unit when the user is initially registered; Daily behavioral data: This data is automatically synchronized via smart devices (such as fitness trackers and mobile apps) or manually entered by the user. This includes daily activity levels (steps, calories burned), dietary records (especially caffeine and alcohol intake), and work schedules (such as commuting times). This type of data is dynamically collected at a frequency of at least once per minute to ensure timeliness. Sleep data: Through smart mattresses, wearable sleep monitoring devices, etc., raw physiological and behavioral signals such as heart rate interval and body movement frequency are automatically collected at night; External treatment plan data: If a user is receiving professional sleep treatment, their treatment plan (such as the Cognitive Behavioral Therapy-I program, medication use plan) can be entered into the system as important reference information.

[0045] Step S2, Data Preprocessing and Normalization: This step is performed by the data preprocessing unit and aims to transform the raw data into high-quality, uniformly formatted model input data. The processing includes: Data cleaning: Handling missing values, such as filling in continuous missing values ​​caused by short-term sensor interruptions using data interpolation methods from nearby time points; Outlier removal: Z-score standardization is used for detection. The mean (μ) and standard deviation (σ) of a certain feature (such as instantaneous heart rate) are calculated, and data points with values ​​exceeding the range of [μ-3σ, μ+3σ] are identified as outliers and removed. Data normalization: This is crucial for generating "normalized input data." Numerical features with different dimensions (such as steps and time) are subjected to min-max normalization, scaling them to the [0,1] interval. Categorical features are one-hot encoded. Ultimately, all multi-source heterogeneous data are transformed into a unified, standardized feature vector that the model can directly process.

[0046] Step S3: Construction, training, and optimization of the sleep analysis model: This step is performed by the model training module.

[0047] Model building and pre-training: The system preferably builds a sleep analysis model based on the Transformer neural network architecture. It uses more than 100,000 desensitized user sleep medical data and their corresponding effective treatment plan cases in the data storage unit as historical datasets to conduct supervised pre-training on the model. The training goal is to enable the model to learn the complex mapping relationship between "user state characteristics" (composed of multi-dimensional data normalized in step S2) and "optimal sleep intervention strategy". Continuous Optimization Mechanism: After the system is put into use, it enters the continuous optimization phase. The key to this invention lies in establishing a feedback-driven closed-loop optimization mechanism. Specifically, the system sets up a feedback data accumulation trigger mechanism (for example, it is automatically triggered once every 100 valid user feedback data points). The optimization process preferably uses a reinforcement learning algorithm (such as Proximal Policy Optimization, PPO) to transform user feedback (step S7 described later) into reward signals. With the goal of maximizing the long-term sleep improvement effect, the model parameters are incrementally updated, allowing the model to continuously evolve as it is used.

[0048] Step S4: Dynamic generation and delivery of personalized sleep plans: This step is executed by the user plan formulation module, which starts working during the suggested time each day (e.g., afternoon): 1. Real-time data integration: The module obtains the user's real-time standardized data for the day from the data preprocessing unit, such as "Today's exercise level: moderate", "Time to leave work: 18:30", "Dinner contains caffeine"; 2. Intelligent reasoning: The integrated real-time feature vector is input into the optimized sleep analysis model, and the model performs millisecond-level forward reasoning; in a preferred embodiment, the reasoning process includes: comparing the data of the day with the user's historical behavior patterns, identifying significant deviation features (such as "today's exercise volume is far below the average level"), and making a comprehensive judgment in combination with the current time context; 3. Solution Generation and Output: The model outputs a structured, personalized sleep plan, which may include, for example: suggested bedtime: 23:15 (with an allowable error range of ±15 minutes); recommended sleep companion content: relaxation meditation audio A; the system ensures that the total response delay from initiating the request to generating the plan does not exceed 10 seconds, and then pushes the plan to the user.

[0049] Step S5: Sleep effect assessment and index quantification: This step is performed by the sleep information statistics module. The user can supplement the actual sleep details of the previous night through the interface of this module the following morning: Input data: actual bedtime (23:05), actual sleep time (23:40), wake-up time (6:45), number of times and duration of waking up during the night, etc.; Indicator Calculation: Based on the above data, the module automatically calculates key sleep quality indicators, including: sleep latency (35 minutes), total sleep time (6 hours and 50 minutes), sleep efficiency (total sleep time / time in bed, for example ≈89%), and time spent in bed. These indicators provide users with a quantitative sleep quality report. Step S6: Proactive Risk Monitoring and Tiered Intervention This step is executed by the data analysis module, which continuously monitors the time-series data of key indicators such as the user's sleep efficiency in the background and runs a pre-set multi-level risk warning logic: Warning Rules: The system defines multiple threshold levels. For example, if sleep efficiency is below 60% for 3 consecutive days, a Level 1 warning is triggered, suggesting "consulting a sleep specialist"; if it is below 70% for 5 consecutive days, a Level 2 warning is triggered, recommending "using specific sleep aid medical devices"; if it is below 80% for 7 consecutive days, a Level 3 warning is triggered, prompting "adjusting daily routines". When the monitoring data meets any of the conditions, the system automatically generates and pushes corresponding graded intervention suggestions to achieve proactive management of sleep health.

[0050] Step S7, Closed-loop feedback and system self-evolution: This step is the core of achieving continuous system intelligence. The objective sleep indicators calculated in step S5 (such as the change in sleep efficiency compared to the previous day) and the subjective user feedback collected in step S6 (such as the "adopt" or "ignore" operation of the warning suggestions) together constitute the feedback data required for optimization. These data are fed back to the model training module (step S3) as training samples for a new round of reinforcement learning. Through this closed loop, the actual usage effects of users are continuously used to adjust and improve the performance of the sleep analysis model, making the system's output suggestions more and more accurate and personalized over time, realizing the transformation from a static tool to an intelligent companion.

[0051] In summary, by organically combining the above steps and modules, this invention constructs a highly adaptive, sustainably evolving, and user-centric intelligent insomnia cognitive management solution, which effectively overcomes the shortcomings of existing technologies, such as rigid recommendations, lack of personalization, and lack of long-term adaptability.

[0052] The advantages of the insomnia cognitive management method and system with intelligent sleep time suggestion of the present invention compared with the prior art are as follows: 1. Breakthrough in data utilization: Through the original four-dimensional data fusion training method of "user basic information + daily behavior data + sleep data + medical treatment plan", it overcomes the shortcomings of existing technologies such as single data dimension and one-sided model profile, and realizes systematic modeling of user sleep ecology.

[0053] 2. Real-time personalized dynamic generation of suggestions: Based on real-time data of the day, customized solutions (such as precise bedtime and sleep companion content) are inferred within seconds through the sleep analysis model, which completely changes the traditional static and universal suggestion push mode and improves the accuracy and timeliness of intervention.

[0054] 3. A continuously optimized intelligent closed loop has been constructed: By automatically converting user feedback data into model training samples through an original iterative mechanism, the system has the ability to continuously learn from actual results. The accuracy of the solution can increase linearly with the usage time, solving the problem of existing systems being "unchanging".

[0055] 4. A proactive health risk management system has been established: Through a multi-level early warning mechanism, it can proactively identify sleep disorder risks and provide tiered guidance from behavioral adjustment to medical intervention, realizing a leap from passive recording to proactive, tiered health management.

[0056] In summary, this invention provides a cognitive management method and system for insomnia with intelligent suggestions for bedtime. The method includes: collecting and preprocessing multi-dimensional user data to obtain standardized input data; training a sleep analysis model based on historical datasets and continuously optimizing the model using user feedback data; inferring from the sleep analysis model based on real-time user data for the day, dynamically generating and outputting a personalized sleep plan including suggested bedtimes; obtaining actual sleep details and calculating sleep quality indicators; conducting continuous risk monitoring based on sleep quality indicators and triggering tiered intervention suggestions; and feeding the sleep quality indicators and user feedback data as training data into the model optimization step to achieve closed-loop optimization. The system includes a data collection module, a data preprocessing unit, a model training module, a user plan formulation module, a sleep information statistics module, a data analysis module, and a data storage unit. This invention solves the problems of rigid suggestions, lack of personalization, and insufficient continuous optimization capabilities in existing technologies, achieving precise, dynamic, and intelligent sleep management.

[0057] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A cognitive management method for insomnia with intelligent suggestions for sleep onset time, characterized in that, At least including: S1. Collect multi-dimensional data of users, including basic user information, daily behavior data, sleep data, and external treatment plan data; S2. Preprocess the collected multi-dimensional data to obtain standardized input data; S3. Train a sleep analysis model based on a historical dataset and optimize the sleep analysis model using user feedback data. The historical dataset contains a large amount of users' sleep medical data and insomnia treatment case studies. S4. Based on the real-time user data in the standardized input data obtained in step S2, the sleep analysis model is invoked to perform inference, dynamically generate and output a personalized sleep plan. S5. Obtain the user's actual sleep details for the previous sleep cycle, and calculate the sleep quality index based on the actual sleep details; S6. Based on the sleep quality index calculated in step S5, continuous monitoring is performed. When the monitoring data reaches the preset risk warning threshold, the corresponding graded intervention suggestion is triggered. S7. The sleep quality index obtained in step S5 and the user feedback data collected in step S6 are used as training data and fed back to step S3 to optimize the parameters of the sleep analysis model and achieve closed-loop optimization of the system.

2. The insomnia cognitive management method with intelligent sleep time suggestion as described in claim 1, characterized in that, In step S3, when training the sleep analysis model, the constructed model input feature vector shall include at least the following four dimensions of associated data: User basic attribute dimensions, including age and body mass index; The daily behavior time series dimension includes time series data on exercise volume and dietary intake events; Historical sleep feedback dimensions include historical sleep efficiency and sleep latency; The external intervention tag dimension includes features of validated treatment protocols that match the user profile.

3. The insomnia cognitive management method with intelligent sleep time suggestion as described in claim 1, characterized in that, In step S7, the training data fed back to step S3 includes: Objective feedback data: Changes in sleep efficiency calculated in step S5; Subjective feedback data: User satisfaction or behavioral feedback indicators of the personalized sleep plan generated in step S4 or the graded intervention suggestions generated in step S6; The sleep analysis model uses both objective and subjective feedback data to construct a reward function for reinforcement learning in order to optimize the model parameters.

4. The insomnia cognitive management method with intelligent sleep time suggestion as described in claim 1, characterized in that, The sleep analysis model is a large-scale sleep model built on the Transformer architecture. The large-scale sleep model is trained on a historical dataset and continuously optimized using user feedback data through a reinforcement learning algorithm. The continuous optimization specifically involves setting up a feedback data accumulation trigger mechanism, which triggers an incremental update of the model parameters every time a predetermined number of valid user feedback data are accumulated.

5. The insomnia cognitive management method with intelligent sleep time suggestion according to any one of claims 1-4, characterized in that, In step S4, the invocation of the sleep analysis model for inference specifically includes: The real-time user data is compared and analyzed with the pre-stored historical user behavior pattern data; Identify deviations in daily behavior relative to historical patterns; Based on the aforementioned deviation characteristics and the current time, the sleep analysis model adjusts the suggested bedtime and the type of sleep companion content in real time.

6. The insomnia cognitive management method with intelligent sleep time suggestion according to any one of claims 1-4, characterized in that, In step S4, the real-time user data includes at least the user's daily exercise volume, diet, and commuting time. The personalized sleep plan includes a suggested bedtime and its allowable error range, as well as recommended sleep companions. The response delay from data input to plan generation does not exceed 10 seconds.

7. The insomnia cognitive management method with intelligent sleep time suggestion according to any one of claims 1-4, characterized in that, In step S5, the actual sleep details include actual bedtime, actual sleep onset time, wake-up time, wake-up time, and nighttime wake-up time; the sleep quality indicators include sleep efficiency, total sleep time, time spent in bed, and time spent before falling asleep.

8. The insomnia cognitive management method with intelligent sleep time suggestion according to any one of claims 1-4, characterized in that, In step S6, the preset risk warning threshold is a multi-level threshold, specifically including: If sleep efficiency falls below the first threshold for the first consecutive scheduled days, a level one warning is triggered and a medical advice is generated. When sleep efficiency falls below the second threshold for the second consecutive predetermined number of days, a level 2 warning is triggered and a medical device recommendation is generated. When sleep efficiency falls below the third threshold for the third consecutive predetermined number of days, a level 3 warning is triggered and a reminder to adjust sleep schedule is generated.

9. The insomnia cognitive management method with intelligent sleep time suggestion according to claim 8, characterized in that, The first predetermined number of days is 3 days, and the first threshold is 60%; the second predetermined number of days is 5 days, and the second threshold is 70%; the third predetermined number of days is 7 days, and the third threshold is 80%.

10. A cognitive management system for insomnia with intelligent sleep time suggestion, used to implement the cognitive management method for insomnia with intelligent sleep time suggestion as described in any one of claims 1-9, characterized in that, The system includes: The data collection module is used to collect multi-dimensional user data; A data preprocessing unit, connected to the data collection module, is used to clean, convert the format of, and remove outliers from the collected data. The model training module is used to build and train the sleep analysis model and continuously optimize the model parameters using feedback data. The user plan formulation module is connected to the data preprocessing unit and the sleep analysis model, and is used to generate a personalized sleep plan based on the user's real-time data by calling the sleep analysis model. The sleep information statistics module is used to obtain actual sleep details supplemented by users and calculate sleep quality indicators; The data analysis module, connected to the sleep information statistics module, is used to monitor sleep quality indicators and generate graded intervention suggestions based on multi-level risk warning thresholds; Data storage unit is used to store user data, model parameters, historical schemes and feedback data; The outputs of the sleep information statistics module and the data analysis module are fed back to the model training module, forming a closed-loop optimization system.

Images