Output method and device of sleep intervention action, computer device and storage medium

Abstract

The present application relates to the technical field of digital medical treatment, and discloses an output method of sleep intervention action, comprising: using a set signal processing mode to perform feature extraction on a time sequence physiological signal generated by a user in a sleep process in a preset time period, to obtain a plurality of physiological signal features; splicing the plurality of physiological signal features with user baseline data to obtain input data, using a pre-trained risk assessment model to predict a sleep risk event of the input data, and outputting probability values of different types of sleep risk events occurring in the user in the preset time period; when the probability value is greater than a preset value, determining a target type corresponding to the sleep risk event occurring in the preset time period, using a pre-trained action prediction model to predict a sleep intervention action of the input data, and outputting a sleep intervention action to be taken for the sleep risk event. The present application can individually assess a sleep risk, and propose a targeted sleep intervention action based on the sleep risk.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of digital medicine, in particular to a sleep intervention action output method and device, computer equipment and a storage medium. BACKGROUND

[0002] Sleep is very important to the human body for ordinary people, and in modern society, work pressure is increasing, and sleep problems are also becoming a growing disturbance. By collecting the individual victory signal degree to assess the risk of sleep, the sleep quality of people can be understood, and through the sleep quality, functions such as disease auxiliary diagnosis, health management, and remote consultation can be realized, and timely sleep intervention can be taken in the presence of sleep risk, which is of great significance to sleep health management and risk prevention.

[0003] In related technologies, a traditional scale evaluation method such as the Pittsburgh Sleep Quality Index Scale can be used to evaluate sleep risk, which defines sleep habits as a problem to the user to be evaluated, and evaluates sleep risk and outputs sleep intervention actions according to the answers to the sleep habit questions. However, the traditional scale evaluation method is usually developed based on a group, and the sleep evaluation method for the group is single and cannot achieve personalized sleep risk evaluation, and cannot propose targeted sleep intervention actions based on sleep risk. SUMMARY

[0004] Therefore, the present application provides a sleep intervention action output method, device, computer equipment and storage medium, which mainly aims to solve the problem that the sleep evaluation method for the group in the prior art is single and cannot achieve personalized sleep risk evaluation, and cannot propose targeted sleep intervention actions based on sleep risk.

[0005] According to one aspect of the present application, a sleep intervention action output method is provided, comprising:

[0006] using a set signal processing method to extract features of a plurality of physiological signals generated by a user during sleep in a preset time period, to obtain a plurality of physiological signal features;

[0007] concatenating the plurality of physiological signal features and user baseline data as input data, using a pre-trained risk assessment model to predict sleep risk events of the input data, and outputting probability values of different types of sleep risk events occurring in the user in the preset time period;

[0008] when the probability value is greater than a preset value, determining a target type corresponding to the sleep risk event occurring in the preset time period, and using a pre-trained action prediction model to predict sleep intervention actions of the input data, and outputting sleep intervention actions to be taken for the target type of sleep risk event.

[0009] According to another aspect of the present application, an output device of a sleep intervention action is provided, comprising:

[0010] an extraction module configured to extract features of a time-series physiological signal generated by a user during sleep within a preset time period using a set signal processing method, to obtain a plurality of physiological signal features;

[0011] a prediction module configured to splice the plurality of physiological signal features with baseline data of the user as input data, and use a pre-trained risk assessment model to predict a sleep risk event of the input data, to output a probability value of different types of sleep risk events occurring within the preset time period;

[0012] an output module configured to determine a target type of a sleep risk event occurring within the preset time period when the probability value is greater than a preset value, and use a pre-trained action prediction model to predict a sleep intervention action of the input data, to output a sleep intervention action to be taken for the target type of sleep risk event.

[0013] According to still another aspect of the present application, a computer device is provided, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the output method of the sleep intervention action when executing the computer program.

[0014] According to yet another aspect of the present application, a computer storage medium is provided, which stores a computer program, and the computer program implements the steps of the output method of the sleep intervention action when executed by a processor.

[0015] By employing the above technical solution, this invention provides a method, apparatus, computer device, and storage medium for outputting sleep intervention actions. It extracts features from temporal physiological signals generated by a user during sleep within a preset time period using a predetermined signal processing method, obtaining various physiological signal features. These features are then concatenated with the user's baseline data as input data. A pre-trained risk assessment model predicts sleep risk events from the input data, outputting probability values ​​for different types of sleep risk events occurring within the preset time period. When the probability value exceeds a preset value, the target type corresponding to the sleep risk event occurring within the preset time period is determined. A pre-trained action prediction model then predicts sleep intervention actions from the input data, outputting the appropriate sleep intervention action for the target type of sleep risk event. Compared to existing technologies that use scale assessment results to output sleep intervention actions, this application combines sleep risk with intervention action prediction, enabling personalized sleep risk assessment to generate intervention actions corresponding to the user's sleep risk in different states, and proposing targeted sleep intervention actions based on sleep risk. Attached Figure Description

[0016] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0017] Figure 1 This is a schematic diagram of an application environment for the output method of sleep intervention actions in one embodiment of the present invention;

[0018] Figure 2 This is a flowchart illustrating a method for outputting sleep intervention actions according to an embodiment of the present invention;

[0019] Figure 3 yes Figure 2 A schematic diagram of a specific implementation method for step S10;

[0020] Figure 4 This is another schematic diagram of the output method of sleep intervention actions in another embodiment of the present invention;

[0021] Figure 5 This is another schematic diagram of the output method of sleep intervention actions in another embodiment of the present invention;

[0022] Figure 6 yes Figure 5 A schematic diagram of a specific implementation of step S70;

[0023] Figure 7 yesFigure 5 A schematic diagram of a specific implementation method for step S80;

[0024] Figure 8 This is a schematic diagram of the process of training an action prediction model in one embodiment of the present invention;

[0025] Figure 9 This is a schematic diagram of the output device for sleep intervention actions in one embodiment of the present invention;

[0026] Figure 10 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention;

[0027] Figure 11 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0028] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0029] The sleep intervention action output method provided in this embodiment of the invention can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The client can receive temporal physiological signals generated by the user during sleep within a preset time period. The server can use a set signal processing method to extract features from the temporal physiological signals generated by the user during sleep within the preset time period. These multiple physiological signal features are then concatenated with the user's baseline data as input data. A pre-trained risk assessment model is used to predict sleep risk events based on the input data, outputting the probability values ​​of different types of sleep risk events occurring within the preset time period. When the probability value is greater than a preset value, the target type corresponding to the sleep risk event occurring within the preset time period is determined. A pre-trained action prediction model is then used to predict sleep intervention actions based on the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event. In this invention, by combining sleep risk with intervention action prediction, personalized sleep risk assessment can be performed to generate intervention actions corresponding to the user's sleep risk in different states, and targeted sleep intervention actions can be proposed based on sleep risk. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The present invention will now be described in detail through specific embodiments.

[0030] Please see Figure 2 As shown, Figure 2 A flowchart illustrating the output method of sleep intervention actions provided in an embodiment of the present invention includes the following steps:

[0031] S10. Use the set signal processing method to extract features of the time-series physiological signals generated by the user during sleep within a preset time period, and obtain a variety of physiological signal features.

[0032] Typically, wearable devices for physiological monitoring mainly include modules such as information acquisition, signal processing, data communication, and application software. They can integrate biosensors, signal acquisition and processing, and data communication modules into everyday wearable objects for the purpose of measuring various physiological indicators or conducting physiological treatments. Common physiological monitoring includes body temperature, respiratory rate, pulse, blood pressure, blood oxygen, blood glucose, electrocardiogram, and electroencephalogram. Here, the time-series physiological signals generated by the user during sleep can be obtained by monitoring the physiological data output by the wearable device.

[0033] The signal processing methods set here can include, but are not limited to, SVM, KNN, Naive Bayes, decision trees, neural networks, etc. Specifically, different types of time-series physiological signals can be analyzed, and the appropriate signal processing method for the time-series physiological signals can be determined based on the analysis results. Then, the corresponding signal processing method can be used to extract features of the time-series physiological signals generated by the user during sleep within a preset time period, resulting in a variety of physiological signal features.

[0034] It is important to understand that different types of temporal physiological signals have different data characteristics in the time and frequency domains. For example, the temporal physiological signals corresponding to physiological data such as electrocardiograms and pulses have periodic characteristics, while other temporal physiological features do not. Specifically, such as... Figure 3 In step S10, the set signal processing method is used to extract features from the temporal physiological signals generated by the user during sleep within a preset time period, resulting in various physiological signal features, including the following steps:

[0035] S11. Acquire various physiological signal data collected by the user during sleep using wearable devices within a preset time period.

[0036] S12. The various physiological signal data are processed into temporal physiological signals generated by the user during sleep within a preset time period.

[0037] S13. Use the set signal processing method to extract features from the time-series physiological signals to obtain various physiological signal features.

[0038] It is understandable that signals acquired through mobile means are susceptible to interference from environmental noise and body movement, resulting in inconsistent signal quality. Discarding interfered signals directly would increase the workload of repeated sampling, while performing data analysis without signal processing would reduce diagnostic accuracy. Therefore, to effectively utilize the physiological signal data acquired by wearable devices, processing can be performed on different types of physiological signal data to improve its resistance to interference. Specifically, regarding the anti-interference capability of heart rate signals, a motion-interference-resistant heart rate detection algorithm can be used. This algorithm first preprocesses the heart rate signal to filter out baseline drift and high-frequency interference, then coarsely processes the signal using adaptive peak dilation and waveform reconstruction techniques, and finally analyzes the signal spectrum using Fast Fourier Transform to calculate the heart rate value. To avoid the impact of motion interference on the accuracy of heart rate detection, an anti-interference strategy based on high-amplitude interference suppression is incorporated. This strategy locates and suppresses motion-related high-frequency amplitude interference by setting thresholds and compressing the signal.

[0039] Specifically, in the process of extracting features from time-series physiological signals using the set signal processing method, considering that different types of time-series physiological signals require different evaluation indicators in measuring sleep risk, we can first determine the evaluation indicators required for different time-series physiological signals in measuring sleep risk, set the signal processing method according to the evaluation indicators, and then use the set signal processing method to extract features from the time-series physiological signals generated by the user during sleep within a preset time period. For each physiological signal feature, we can extract physiological signal features that can reflect the corresponding evaluation indicators.

[0040] S20. The various physiological signal features are concatenated with the user's baseline data as input data. A pre-trained risk assessment model is used to predict sleep risk events from the input data, and the probability values ​​of different types of sleep risk events occurring within a preset time period are output.

[0041] The baseline data consists of user attribute data, such as age, gender, occupation, and underlying medical conditions. User attribute data plays a crucial role in measuring sleep risk. Generally, under the same physiological signal characteristics, older users have a higher probability of experiencing sleep risk events, and shorter sleep duration due to their occupation also increases the probability of such events. Sleep risk events can include events that may cause sleep apnea or other sleep disorders. In practical applications, the pre-trained risk assessment model can be applied to the field of medical diagnosis. Since sleep risk events can be used to assess a user's sleep quality, the probability values ​​of different types of sleep risk events occurring within a preset time period can be used to assist in medical diagnosis.

[0042] It's important to understand that the risk assessment model is trained using a machine learning model. The entire prediction process involves determining whether a sleep risk event will occur—a binary variable. After building the risk assessment model, it can be directly called to output the probability values ​​of different types of sleep risk events that a user will experience within a preset time period. Specifically, for example... Figure 4 As shown, before step S20, that is, before the various physiological signal features are concatenated with the user's baseline data as input data, and before the pre-trained risk assessment model is used to predict sleep risk events on the input data and outputs the probability values ​​of different types of sleep risk events occurring within a preset time period, the following steps are also included:

[0043] S40. The multiple physiological signal features are spliced ​​together with the user baseline data to form the first sample data, and a machine learning model is called to train the first sample data.

[0044] S50. Pre-set different types of sleep risk events. During the training process of the first sample data, use the binary variable of the first sample data that has different types of sleep risk events as the risk prediction variable. When it is determined during the training process that the risk prediction variable meets the first iteration convergence condition, determine the trained machine learning model as the risk assessment model.

[0045] The first iteration convergence condition can be achieved by setting a loss function. Specifically, the loss function can be used to calculate the loss value formed by the output value and the sleep risk event marked by the physiological signal feature multiple times. If the loss value does not meet the iteration stopping condition, the model parameters corresponding to the machine learning model are adjusted until the output value of the machine learning model meets the first iteration convergence condition, and the trained machine learning model is determined as the risk assessment model.

[0046] In this step, the sleep risk state of different sleep types can be determined based on whether the physiological signal characteristics in the first sample data occur within a set time. If it occurs, the first sample data is determined to have experienced different types of sleep risk assessment events. For example, whether a sleep apnea event has occurred can be judged by whether there is no breathing within 10 seconds. If the physiological signal characteristics in the first sample data indicate that there is no breathing within 10 seconds, then a sleep apnea event is determined to have occurred.

[0047] S30. When the probability value is greater than a preset value, determine the target type corresponding to the sleep risk event that occurred within a preset time period, and use a pre-trained action prediction model to predict sleep intervention actions on the input data, and output the sleep intervention actions to be taken for the sleep risk event of the target type.

[0048] The higher the risk value, the higher the probability that the user will experience different types of sleep risk events within the preset time period. Only when a sleep risk event occurs should corresponding sleep intervention actions be taken. Here, when the probability value is greater than the preset value, it can be determined that the user has experienced a sleep risk event within the preset time period. When the probability value is greater than the preset value, it can be determined that a target type of sleep risk event has occurred within the preset time period.

[0049] In this step, sleep intervention actions may include, but are not limited to, sound wake-up and / or vibration wake-up. Specifically, sound wake-up may include different types of music and / or voice, and vibration wake-up may include different vibration patterns.

[0050] It's important to understand that the action prediction model is trained using a reinforcement learning model. The entire prediction process determines whether the action prediction variable has an effect on sleep risk time. After building the action prediction model, it can be directly called to output sleep intervention actions to be taken for sleep risk time. Specifically, for example... Figure 5As shown, before step S30, that is, before using a pre-trained action prediction model to predict sleep intervention actions from the input data and outputting the sleep intervention actions to be taken for the target type of sleep risk event, the following steps are also included:

[0051] S60. The probability values ​​of different types of sleep risk events that the user will experience within a preset time period are concatenated with the input data to form the second sample data. The reinforcement learning model is then called to train the second sample data.

[0052] S70. During the training process of the second sample data, multiple sleep intervention actions are pre-set, and after each sleep intervention action is used on the input data, the range of probability values ​​of different types of sleep risk events occurring in the user in the future time period is determined.

[0053] S80. Generate prediction variables based on the range of the probability values. When the action prediction variables meet the second iteration convergence condition during the training process, determine the trained reinforcement learning model as the action prediction model.

[0054] Specifically, such as Figure 6 As shown, in step S70, that is, during the training process of the second sample data, multiple sleep intervention actions are pre-set, and after using each sleep intervention action on the input data, the range of probability values ​​for different types of sleep risk events occurring in the user within a future time period is determined, including the following steps:

[0055] S71. During the training process of the second sample data, the input data in the second sample data is divided into multiple physiological feature data according to time periods.

[0056] S72. Pre-set multiple sleep intervention actions, and after applying each sleep intervention action to the physiological characteristic data of the current time period, obtain the probability values ​​of the predicted physiological characteristic data of the future time period on different types of sleep risk events.

[0057] S73. Based on the probability values ​​of physiological characteristic data for the future time period on different types of sleep risk events, determine the range of values ​​for the probability values ​​of different types of sleep risk events that the user will experience in the future time period.

[0058] Specifically, such as Figure 7 As shown, in step S80, that is, generating prediction variables based on the range of the probability values, and determining the trained reinforcement learning model as the action prediction model after the action prediction variables meet the second iteration convergence condition during the training process, the following steps are included:

[0059] S81. Based on the range of the probability value, learning resources are distributed according to the contribution of each sleep intervention action to the sleep intervention effect.

[0060] S82. The sum of learning resources accumulated over the future time period is used as the action prediction variable. When the action prediction variable meets the second iteration convergence condition during the training process, the trained reinforcement learning model is determined as the action prediction model.

[0061] The specific training process of the action prediction model is as follows: First, sleep time-series features are input into the machine learning model to train it. During training, the prediction target is whether sleep apnea occurs, and the probability of sleep apnea events is output. Then, the state of the sleep time-series features at each time point within a certain period is recorded. The state at each time point can be the feature state at a set time point, such as the average, maximum, or minimum value of the sleep time-series features per minute, which can be customized. The reinforcement training model is then called, using the probability value output by the machine learning model and the sleep time-series features as inputs. During the initial training, the first state is randomly selected based on the first state at each time point within the at least certain period. The first sleep intervention action corresponding to the state is used. Then, after using the sleep intervention action, the probability value of the sleep apnea event output by the machine learning model is recorded. At the same time, the learning resources that should be distributed in the next time after using the first sleep intervention action are determined, and the learning resources that should be distributed in future time are accumulated to obtain the total learning resources that should be distributed for the first sleep intervention action. Here, the training objective of reinforcement learning is to continuously maximize the total learning resources. When the total learning resources are maximized, the training process for the first sleep intervention action is completed. The learning resources that should be distributed for other sleep intervention actions can be determined and accumulated in a corresponding way to complete the training process for other sleep intervention actions. The reinforcement learning model obtained by training is used as the action prediction model.

[0062] Considering the magnitude of the effect after using sleep intervention actions, the probability value can be divided into multiple ranges. Learning resources should be pre-set for each range corresponding to different probability values. For example, if the probability of a user experiencing a sleep risk event after using a sleep intervention action is >= 0.8, then the learning resource to be distributed for that sleep intervention action is -10 points; if the probability is >= 0.5 and < 0.8, then the learning resource to be distributed for that sleep intervention action is -1 point; and if the probability is < 0.5, then the learning resource to be distributed for that sleep intervention action is 1 point. Then, the sum of learning resources accumulated over future time periods can be used as the action prediction variable. Each sleep intervention action is trained iteratively. When the action prediction variable is maximized during the training process for each sleep intervention action, the reinforcement learning model trained on multiple sleep intervention actions is summarized as the action prediction model.

[0063] In practical applications, the specific process of training an action prediction model is as follows: Figure 8 As shown in the diagram, referring to the reinforcement learning process, the first state S1 represents the state of the sleep risk event at the first moment. The reinforcement learning model to be trained can be invoked to output the first sleep intervention action (Action) based on the first state, and the corresponding first learning resource (Reward) can be determined. Since the sleep risk event transitions from the first state S1 to the second state S2, which represents the state of the sleep risk event at the second moment, the reinforcement learning model to be trained can be invoked to output the second sleep intervention action (Action) based on the second state, and the corresponding second learning resource (Reward) can be determined. The first and second learning resource rewards are added to obtain the sum of the learning resources R. When R is the maximum R, a pre-trained reinforcement learning model is obtained. If certain behaviors cause R to decrease, these behaviors will be avoided in the next training iteration.

[0064] This embodiment provides a method for outputting sleep intervention actions. It extracts features from the temporal physiological signals generated by a user during sleep within a preset time period using a pre-defined signal processing method, obtaining various physiological signal features. These features are then concatenated with the user's baseline data as input data. A pre-trained risk assessment model predicts sleep risk events from the input data, outputting the probability values ​​of different types of sleep risk events occurring within the preset time period. When the probability value is greater than a preset value, the target type corresponding to the sleep risk event within the preset time period is determined. A pre-trained action prediction model then predicts sleep intervention actions from the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event. Compared to existing technologies that use scale assessment results to output sleep intervention actions, this application combines sleep risk with intervention action prediction, enabling personalized assessment of sleep risk to generate intervention actions corresponding to the user's sleep risk in different states, and proposing targeted sleep intervention actions based on sleep risk.

[0065] In one embodiment, an output device for sleep intervention actions is provided, which corresponds one-to-one with the output methods for sleep intervention actions described in the above embodiments. For example... Figure 9 As shown, the output device for this sleep intervention includes: an extraction module 101, a prediction module 102, and an output module 103. Detailed descriptions of each functional module are as follows:

[0066] The extraction module 101 can be used to extract features of the temporal physiological signals generated by the user during sleep within a preset time period using a set signal processing method, and obtain a variety of physiological signal features.

[0067] The prediction module 102 can be used to concatenate the various physiological signal features with the user's baseline data as input data, use a pre-trained risk assessment model to predict sleep risk events on the input data, and output the probability values ​​of different types of sleep risk events that the user will experience within a preset time period.

[0068] The output module 103 can be used to determine the target type corresponding to the sleep risk event that occurs within a preset time period when the probability value is greater than a preset value, and use a pre-trained action prediction model to predict sleep intervention actions on the input data, and output the sleep intervention actions that should be taken for the sleep risk event of the target type.

[0069] In one embodiment, the extraction module 101 can be specifically used to acquire various physiological signal data collected by the user during sleep using a wearable device within a preset time period; process the various physiological signal data into temporal physiological signals generated by the user during sleep within the preset time period; and extract features from the temporal physiological signals using a set signal processing method to obtain various physiological signal features.

[0070] In one embodiment, the device further includes:

[0071] The first training module can be used to concatenate the various physiological signal features with user baseline data as the first sample data, and call a machine learning model to train the first sample data.

[0072] The first determining module can be used to pre-set different types of sleep risk events. During the training process of the first sample data, the binary variable of the first sample data having different types of sleep risk events is used as the risk prediction variable. When the risk prediction variable meets the first iteration convergence condition during the training process, the trained machine learning model is determined as the risk assessment model.

[0073] In one embodiment, the device further includes:

[0074] The second training module can be used to concatenate the probability values ​​of different types of sleep risk events that the user will experience within a preset time period with the input data to form the second sample data, and then call the reinforcement learning model to train the second sample data.

[0075] The setting module can be used to pre-set multiple sleep intervention actions during the training process of the second sample data, and after using each sleep intervention action on the input data, determine the range of probability values ​​of different types of sleep risk events that the user will experience in the future time period.

[0076] The second determining module can be used to generate predictive variables based on the range of values ​​of the probability values. When the action predictive variables meet the second iteration convergence condition during the training process, the trained reinforcement learning model is determined as the action prediction model.

[0077] In one embodiment, the setting module can be specifically used to, during the training process of the second sample data, divide the input data in the second sample data into multiple physiological feature data according to time periods; pre-set multiple sleep intervention actions; after applying each sleep intervention action to the physiological feature data of the current time period, obtain the probability values ​​of the predicted physiological feature data of the future time period on different types of sleep risk events; and determine the range of values ​​of the probability values ​​of the user's occurrence of different types of sleep risk events in the future time period based on the probability values ​​of the physiological feature data of the future time period on different types of sleep risk events.

[0078] In one embodiment, the second determining module can be specifically used to, during the training process of the second sample data, divide the input data in the second sample data into multiple physiological feature data according to time periods; pre-set multiple sleep intervention actions, and after applying each sleep intervention action to the physiological feature data of the current time period, obtain the probability values ​​of the predicted physiological feature data of the future time period on different types of sleep risk events; and determine the range of values ​​of the probability values ​​of the user's occurrence of different types of sleep risk events in the future time period based on the probability values ​​of the physiological feature data of the future time period on different types of sleep risk events.

[0079] In one embodiment, the device further includes:

[0080] The segmentation module can be used to divide the probability value into multiple value ranges before distributing learning resources based on the contribution of each sleep intervention action to the sleep intervention effect according to the value range in which the probability value is located, and to pre-set the learning resources that should be distributed for different probability values ​​corresponding to different value ranges.

[0081] The second determining module can also be used to use the sum of learning resources accumulated over a future time period as the action prediction variable, iterate through and train each sleep intervention action, and when the action prediction variable reaches its maximum during the training process for each sleep intervention action, summarize the reinforcement learning model trained on multiple sleep intervention actions as the action prediction model.

[0082] This embodiment provides a sleep intervention action output device. It extracts features from the temporal physiological signals generated by a user during sleep within a preset time period using a set signal processing method, obtaining various physiological signal features. These features are then concatenated with the user's baseline data as input data. A pre-trained risk assessment model predicts sleep risk events from the input data, outputting the probability values ​​of different types of sleep risk events occurring within the preset time period. When the probability value is greater than a preset value, the target type corresponding to the sleep risk event within the preset time period is determined. A pre-trained action prediction model then predicts sleep intervention actions from the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event. Compared to existing technologies that use scale assessment results to output sleep intervention actions, this application combines sleep risk with intervention action prediction, enabling personalized assessment of sleep risk to generate intervention actions corresponding to the user's sleep risk in different states, and proposing targeted sleep intervention actions based on sleep risk.

[0083] Specific limitations regarding the output device for sleep intervention actions can be found in the limitations on the output method of sleep intervention actions described above, and will not be repeated here. Each module in the aforementioned output device for sleep intervention actions can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0084] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a sleep intervention action output method on the server side.

[0085] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 11As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements the functions or steps of a sleep intervention action output method on the client side.

[0086] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0087] The set signal processing method is used to extract features of the temporal physiological signals generated by the user during sleep within a preset time period, resulting in a variety of physiological signal features;

[0088] The various physiological signal features are concatenated with the user's baseline data as input data. A pre-trained risk assessment model is used to predict sleep risk events based on the input data, and the probability values ​​of different types of sleep risk events occurring in the user within a preset time period are output.

[0089] When the probability value is greater than a preset value, the target type corresponding to the sleep risk event occurring within a preset time period is determined, and a pre-trained action prediction model is used to predict sleep intervention actions based on the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event.

[0090] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0091] The set signal processing method is used to extract features of the temporal physiological signals generated by the user during sleep within a preset time period, resulting in a variety of physiological signal features;

[0092] The various physiological signal features are concatenated with the user's baseline data as input data. A pre-trained risk assessment model is used to predict sleep risk events based on the input data, and the probability values ​​of different types of sleep risk events occurring in the user within a preset time period are output.

[0093] When the probability value is greater than a preset value, the target type corresponding to the sleep risk event occurring within a preset time period is determined, and a pre-trained action prediction model is used to predict sleep intervention actions based on the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event.

[0094] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0097] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for outputting sleep intervention actions, characterized in that, The method includes: The set signal processing method is used to extract features of the temporal physiological signals generated by the user during sleep within a preset time period, resulting in a variety of physiological signal features; The various physiological signal features are concatenated with the user's baseline data as input data. A pre-trained risk assessment model is used to predict sleep risk events on the input data, and the output is the probability value of the user experiencing different types of sleep risk events within a preset time period. The risk assessment model is trained using a machine learning model. The prediction process is a binary variable that determines whether a sleep risk event will occur. After constructing the risk assessment model, the probability value of the user experiencing different types of sleep risk events within a preset time period is output by calling the risk assessment model. When the probability value of the user experiencing different types of sleep risk events within a preset time period is greater than a preset value, the target type corresponding to the sleep risk event occurring within the preset time period is determined, and a pre-trained action prediction model is used to predict sleep intervention actions on the input data, outputting the sleep intervention actions to be taken for the target type of sleep risk event. The method further includes: concatenating the probability values ​​of different types of sleep risk events occurring in the user within a preset time period with the input data to obtain second sample data; calling a reinforcement learning model to train the second sample data; during the training of the second sample data, pre-setting multiple sleep intervention actions, and after using each sleep intervention action on the input data, determining the range of probability values ​​of different types of sleep risk events occurring in the user within a future time period; generating action prediction variables based on the range of probability values ​​of different types of sleep risk events occurring in the user within a future time period; and determining the trained reinforcement learning model as the action prediction model after the action prediction variables meet the second iteration convergence condition during the training process.

2. The method according to claim 1, characterized in that, The process involves using a pre-defined signal processing method to extract features from the temporal physiological signals generated by the user during sleep within a preset time period, resulting in various physiological signal features, including: Acquire various physiological signal data collected by the user during sleep using wearable devices within a preset time period; The various physiological signal data are processed into temporal physiological signals generated by the user during sleep within a preset time period; The time-series physiological signals are subjected to feature extraction using a set signal processing method to obtain various physiological signal features.

3. The method according to claim 1, characterized in that, The method further includes: The various physiological signal features are spliced ​​together with the user baseline data to form the first sample data, and a machine learning model is called to train the first sample data. Different types of sleep risk events are pre-set. During the training process of the first sample data, the binary variable of the first sample data that has different types of sleep risk events is used as the risk prediction variable. When the risk prediction variable meets the first iteration convergence condition during the training process, the trained machine learning model is determined as the risk assessment model.

4. The method according to claim 1, characterized in that, During the training process on the second sample data, multiple sleep intervention actions are pre-set, and after applying each sleep intervention action to the input data, the range of probability values ​​for different types of sleep risk events occurring in the user within a future time period is determined, including: During the training process of the second sample data, the input data in the second sample data is divided into multiple physiological feature data according to time periods; Multiple sleep intervention actions are pre-set. After applying each sleep intervention action to the physiological characteristic data of the current time period, the probability values ​​of the predicted physiological characteristic data of the future time period on different types of sleep risk events are obtained. Based on the probability values ​​of physiological characteristic data for the future time period on different types of sleep risk events, the range of probability values ​​for the user to experience different types of sleep risk events in the future time period is determined.

5. The method according to claim 1, characterized in that, The process of generating action prediction variables based on the probability range of different types of sleep risk events occurring in the user within a future time period, and determining the trained reinforcement learning model as the action prediction model after the action prediction variables meet the second iteration convergence condition during the training process, includes: Learning resources are distributed based on the range of probability values ​​for different types of sleep risk events that the user will experience in the future time period, according to the contribution of each sleep intervention action to the sleep intervention effect. The sum of learning resources accumulated over a future time period is used as the action prediction variable. When the action prediction variable meets the second iteration convergence condition during the training process, the trained reinforcement learning model is determined as the action prediction model.

6. The method according to claim 5, characterized in that, Before distributing learning resources based on the probability range of different types of sleep risk events occurring in the user within a future time period, and based on the contribution of each sleep intervention action to the sleep intervention effect, the method further includes: Divide the probability values ​​of different types of sleep risk events that users will experience in the future into multiple ranges, and pre-set the learning resources that should be distributed for different ranges. The step of using the sum of learning resources accumulated over future time periods as the action prediction variable, and determining the trained reinforcement learning model as the action prediction model after determining that the action prediction variable meets the second iteration convergence condition during training, includes: The sum of learning resources accumulated over a future time period is used as the action prediction variable. Each sleep intervention action is trained iteratively. When the action prediction variable is maximized during the training process for each sleep intervention action, the reinforcement learning model trained on multiple sleep intervention actions is summarized as the action prediction model.

7. An output device for sleep intervention actions, characterized in that, The device includes: The extraction module is used to extract features from the temporal physiological signals generated by the user during sleep within a preset time period using a set signal processing method, and obtain various physiological signal features. The prediction module is used to concatenate the various physiological signal features with the user's baseline data as input data, and use a pre-trained risk assessment model to predict sleep risk events on the input data. The output is the probability value of the user experiencing different types of sleep risk events within a preset time period. The risk assessment model is trained using a machine learning model. The prediction process is a binary variable that determines whether a sleep risk event will occur. After constructing the risk assessment model, the module outputs the probability value of the user experiencing different types of sleep risk events within a preset time period by calling the risk assessment model. The output module is used to determine the target type of sleep risk event that occurred within the preset time period when the probability value of the user experiencing different types of sleep risk events within the preset time period is greater than a preset value, and to use a pre-trained action prediction model to predict sleep intervention actions for the input data, and output the sleep intervention actions that should be taken for the target type of sleep risk event. The device further includes: a second training module, used to concatenate the probability values ​​of different types of sleep risk events occurring in the user within a preset time period with the input data to obtain second sample data, and call a reinforcement learning model to train the second sample data; a setting module, used to pre-set multiple sleep intervention actions during the training process of the second sample data, and after using each sleep intervention action on the input data, determine the range of probability values ​​of different types of sleep risk events occurring in the user within a future time period; and a second determination module, used to generate action prediction variables based on the range of probability values ​​of different types of sleep risk events occurring in the user within a future time period, and when the action prediction variables meet the second iterative convergence condition during the training process, determine the trained reinforcement learning model as the action prediction model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images