Wearable device sleep data processing method and system supporting low power consumption

Abstract

The invention discloses a wearable device sleep data processing method and system supporting low power consumption, and relates to the related technical field of data processing.The method comprises the steps that sleep stages are divided, monitoring sampling frequencies and sensing sources are distributed to all the sleep stages, and monitoring rules are set. Sample driving is used as a training mode, a low-power-consumption module is constructed, and the low-power-consumption module is arranged in a processor of the wearable device and comprises a front data interface and a lightweight data model. And performing mode management on peripheral equipment of the processor based on the monitoring rule, and monitoring and acquiring multi-mode sensing data. And based on a front data interface, performing multi-modal fusion and data lightweight compression on the multi-modal sensing data, transferring to a lightweight data model for sleep state evaluation, and performing time sequence integration to output a sleep map. The technical problems that in the prior art, a sleep monitoring device is high in power consumption, and it is difficult to accurately monitor the sleep state of a user in a low-power-consumption state are solved.

Description

A method and system for processing sleep data in wearable devices that supports low power consumption. Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method and system for processing sleep data of wearable devices that supports low power consumption. Background Technology

[0002] With the widespread adoption of smart wearable devices, sleep monitoring technology has become an important tool for improving health management. Traditional sleep monitoring methods mostly rely on cumbersome and high-power-consuming devices, leading to problems such as high power consumption and complex data processing during long-term use. To improve the user experience and extend the lifespan of wearable devices, how to efficiently perform sleep monitoring under low-power conditions has become a key issue in technological development.

[0003] Therefore, existing sleep monitoring devices have high power consumption, making it difficult to accurately monitor the user's sleep state in a low-power state. Summary of the Invention

[0004] This application provides a method and system for processing sleep data in wearable devices that supports low power consumption, solving the technical problem that existing sleep monitoring devices have high power consumption and are difficult to accurately monitor the user's sleep state under low power conditions. By dynamically dividing sleep stages, allocating different sensors and sampling frequencies, and combining lightweight data models and multimodal fusion data compression technology, efficient sleep state monitoring is achieved under low power conditions, while improving the system's response speed and monitoring accuracy.

[0005] This application provides a low-power wearable device sleep data processing method, the method comprising: dividing sleep stages, allocating monitoring sampling frequencies and sensor sources to each sleep stage, and setting monitoring rules, wherein sleep stages are divided based on sleep depth; constructing a low-power module using a sample-driven training method, the low-power module being built into the processor of the wearable device, including a front-end data interface and a lightweight data model; performing mode management on peripheral devices of the processor based on the monitoring rules, monitoring and acquiring multimodal sensor data; performing multimodal fusion and lightweight data compression on the multimodal sensor data based on the front-end data interface, transferring the data to the lightweight data model for sleep state assessment, and outputting a sleep map by time-series integration, wherein compression is performed by extracting single-point data and data activity.

[0006] In the implementation, based on the monitoring rules, mode management is performed on the processor's peripheral devices, including: introducing a power management module, which performs mode switching of the processor's peripheral devices, including working mode and standby mode; identifying the monitoring rules and writing I / O control commands, wherein the I / O control commands identify the device code and mode type; and performing mode management on the peripheral devices based on the I / O control commands, wherein mode management includes a set of working modes for peripheral devices and two sets of standby modes for peripheral devices.

[0007] In the implementation, the low-power module includes a lightweight data model, comprising: calling sleep state evaluation records, integrating and determining training samples, wherein the training samples include input-multi-level intermediate evaluation-output; supervising the training of an initial data model based on the training samples; performing lightweight processing on the initial data model, and validating it based on the training samples to determine the lightweight data model, wherein the lightweight processing method is the reduction of the number of network layers and parameter types, and the reduction is constrained by a preset influence degree.

[0008] In the implementation method, the multimodal sensing data is fused and compressed in a lightweight manner, and then transferred to the lightweight data model for sleep state assessment. This includes: performing sensing acquisition in the working mode of a set of peripheral devices to determine the multimodal sensing data; performing multimodal fusion and compression on the multimodal sensing data to determine lightweight monitoring data; and transferring the lightweight monitoring data to the lightweight data model to assess and determine the sleep state, wherein the sleep state is identified by a collection timestamp.

[0009] In the implementation method, data compression is performed to determine lightweight monitoring data, including: setting a compression frame rate; based on the compression frame rate, performing frame rate extraction and static feature recognition on the modal fusion data to determine single-point data; identifying dynamic features between frame rates to determine data activity; integrating the single-point data and the data activity in positive time sequence to obtain the lightweight monitoring data; performing analog-to-digital conversion on the lightweight monitoring data and transmitting it to the lightweight data model.

[0010] In the implementation, the time-series integrated output sleep map includes: drawing a sleep map with the phase and sleep state as the axis; identifying the collection timestamp of the output sleep state and updating the sleep map; generating sleep warning information based on the abnormal state determination of the sleep state and the abnormal trend determination of the state change of the sleep map; and the wearable device responding to the sleep warning information to issue an abnormal alarm.

[0011] In the implementation method, after outputting the sleep map, the following steps are taken: based on the sleep map, a new monitoring node is introduced, wherein the new monitoring node is a sampling time node and a node sensor source inserted on the basis of the monitoring rules, and the node sensor source is at least one of the sensor sources; and based on the new monitoring node, the user's sleep monitoring and management are performed.

[0012] This application also provides a low-power wearable device sleep data processing system, comprising: a stage segmentation module, used to divide sleep stages, allocate monitoring sampling frequencies and sensor sources to each sleep stage, and set monitoring rules, wherein sleep stages are divided based on sleep depth; a model building module, used to build a low-power module using sample-driven training, wherein the low-power module is built into the processor of the wearable device, including a front-end data interface and a lightweight data model; a sensor data acquisition module, used to perform mode management on the peripheral devices of the processor based on the monitoring rules, monitor and acquire multimodal sensor data; and a state evaluation module, used to perform multimodal fusion and lightweight data compression on the multimodal sensor data based on the front-end data interface, transfer the data to the lightweight data model for sleep state evaluation, and output a sleep map through time-series integration, wherein compression is performed by extracting single-point data and data activity.

[0013] This application proposes a low-power wearable device sleep data processing method and system. The method involves dividing sleep into stages, allocating monitoring sampling frequencies and sensor sources to each stage, and setting monitoring rules. Sleep stages are divided based on sleep depth. A low-power module is constructed using a sample-driven training method. This low-power module is built into the wearable device's processor and includes a front-end data interface and a lightweight data model. Based on the monitoring rules, the peripheral devices of the processor are managed to monitor and acquire multimodal sensor data. Based on the front-end data interface, the multimodal sensor data undergoes multimodal fusion and lightweight data compression, and is then fed into the lightweight data model for sleep state assessment. A time-series integrated sleep map is output, with compression methods focusing on extracting single-point data and data activity. This method solves the technical problem of high power consumption in existing sleep monitoring devices, making it difficult to accurately monitor the user's sleep state under low-power conditions. By dynamically dividing sleep stages, allocating different sensors and sampling frequencies, and combining a lightweight data model with multimodal fusion and data compression technology, efficient sleep state monitoring under low-power conditions is achieved, while simultaneously improving the system's response speed and monitoring accuracy. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 is a schematic flowchart of a low-power wearable device sleep data processing method provided in an embodiment of this application; Figure 2 is a schematic diagram of a low-power wearable device sleep data processing system provided in an embodiment of this application.

[0016] Figure labeling: Phase division module 11, model building module 12, sensor data acquisition module 13, state assessment module 14. Detailed Implementation

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0020] This application provides a method and system for low-power wearable device sleep data processing, as shown in Figure 1. The method includes: dividing sleep into stages, allocating monitoring sampling frequencies and sensor sources to each sleep stage, and setting monitoring rules, wherein sleep stages are divided based on sleep depth. A low-power module is constructed using a sample-driven training method. The low-power module is built into the processor of the wearable device and includes a front-end data interface and a lightweight data model.

[0021] Sleep stages are segmented by analyzing sleep depth. Sleep stages generally include wakefulness, light sleep, deep sleep, and REM sleep. Based on the characteristics of each sleep stage, monitoring sampling frequencies and sensor sources are assigned. For example, during deep sleep, body movement is less frequent, heart rate is lower, and the requirement for detailed data is lower, so the sensor sampling frequency is appropriately reduced to save power. For REM sleep, higher frequency data acquisition is needed. During light sleep, body movement is frequent, and heart rate is higher. The sensor sources are the types of sensors included in the wearable device, including: accelerometers, heart rate sensors, temperature sensors, and electroencephalogram (EEG) sensors. Different sensor sources are used for different sleep depths. For example, during deep sleep, body temperature usually drops slightly, so a temperature sensor can be used as an auxiliary sensor source to determine the deep sleep stage. Subsequently, sample data from the sleep state assessment records are collected, and a low-power module is constructed using sample-driven training. The low-power module is built into the processor of the wearable device and includes a front-end data interface and a lightweight data model. The front-end data interface is used to acquire sensor data collected in real time by the wearable device and input it into the lightweight data model to perform sleep state assessment.

[0022] The method provided in this application further includes: calling sleep state assessment records and integrating them to determine training samples, wherein the training samples include input-multi-level intermediate assessment-output. Based on the training samples, supervised training of an initialization data model is performed. The initialization data model is then subjected to lightweight processing and validated based on the training samples to determine the lightweight data model, wherein the lightweight processing is constrained by reducing the number of network layers and parameter types.

[0023] The low-power module includes a lightweight data model, comprising: accessing sleep state assessment records, which are data recorded by the user during historical sleep, and integrating these records to determine training samples. The training samples consist of input, multi-level intermediate evaluation, and output. The input is multimodal data provided by various sensors, such as time-series heart rate data and temperature data. The multi-level intermediate evaluation data involves the model first extracting preliminary features (such as body motion frequency and heart rate fluctuations) from the input data during multi-level processing, and then evaluating the initial stages of sleep using a series of algorithms. For example, the distinction between light and deep sleep can be achieved using the data fluctuation range of an accelerometer. The output data is the final determined sleep state result based on the intermediate evaluation results. Based on the training samples, a supervised training initialization data model is performed, which is constructed based on a neural network model. Subsequently, to ensure efficient operation on low-power devices, the initialization data model needs to be lightweighted. The goal of this lightweighting is to reduce the computational complexity of the model, enabling it to run in real-time on wearable devices with limited hardware resources without causing excessive power consumption and latency. By removing unnecessary layers from the original deep learning model, the model's complexity and number of parameters are reduced, resulting in a lightweight data model. The accuracy of the lightweight model's output is evaluated using labeled data from both the validation and training sets. If the accuracy meets the requirements, the lightweight data model is validated and recognized as a successful model.

[0024] Based on the monitoring rules, the peripheral devices of the processor are managed in a specific mode to monitor and acquire multimodal sensor data. Based on the front-end data interface, the multimodal sensor data is fused and compressed using a lightweight data simplification method. The data is then transferred to the lightweight data model for sleep state assessment, and a time-series integrated sleep map is output. The compression method uses the extraction of single-point data and data activity.

[0025] Based on the monitoring rules, the peripheral devices of the processor are managed in different modes. The operation of peripheral devices requiring monitoring is controlled, while peripheral devices not requiring monitoring are shut down. Simultaneously, multimodal sensor data is monitored and acquired, including sensor data collected by the enabled peripheral devices. Finally, data processing is performed based on the front-end data interface to complete the multimodal fusion and lightweight data compression of the multimodal sensor data. Compression is achieved by extracting single-point data and data activity. The processed data is then fed into the lightweight data model for sleep state assessment, obtaining the output sleep state and timestamp, and outputting a sleep map through time-series integration. This solves the technical problem of high power consumption in existing sleep monitoring devices, making it difficult to accurately monitor the user's sleep state under low power conditions. By dynamically dividing sleep stages, allocating different sensors and sampling frequencies, and combining a lightweight data model and multimodal fusion data compression technology, efficient sleep state monitoring under low power conditions is achieved, while simultaneously improving the system's response speed and monitoring accuracy.

[0026] The method provided in this application embodiment further includes: introducing a power management module, which performs mode switching of peripheral devices of the processor, the switching mode including working mode and standby mode. The monitoring rules are identified, and I / O control commands are written, the I / O control commands identifying device codes and mode types. Based on the I / O control commands, mode management is performed on the peripheral devices, wherein mode management includes a set of working modes for one set of peripheral devices and two sets of standby modes for two sets of peripheral devices.

[0027] Based on the monitoring rules, mode management is performed on the processor's peripheral devices, including: introducing a power management module, which is responsible for dynamically adjusting the working state of each peripheral device according to the current needs of the device. In sleep monitoring devices, especially to extend battery life, it is necessary to automatically switch the device's working mode according to the user's different sleep stages. The peripheral devices include hardware such as sensors, which are switched between working and standby modes based on needs. Mode switching is mainly divided into two types: working mode and standby mode. Subsequently, the monitoring rules are identified, which include the sampling frequency and corresponding sensor source selection for different sleep stages. I / O control commands are written based on the monitoring rules, and the I / O control commands are identified by device codes and mode types. The device code identifies the specific hardware device that needs to switch modes (such as heart rate sensors, temperature sensors, etc.), and the mode type identifies the target working mode (such as working mode or standby mode). Based on the I / O control commands, mode management is performed on the peripheral devices, which includes the working modes of one set of peripheral devices and the standby modes of two sets of peripheral devices. When peripheral devices are in operating mode, sensors such as accelerometers and heart rate monitoring modules continuously run, collecting data and transmitting it to the processor for further processing. The processor then analyzes the user's sleep state based on algorithms and adjusts the modes of other peripheral devices accordingly. When the device enters standby mode, the operating frequency and sampling rate of peripheral devices are significantly reduced, or some sensor modules may even be completely shut down to reduce system power consumption.

[0028] The method provided in this application embodiment further includes: performing sensing acquisition in the working mode of a set of peripheral devices to determine the multimodal sensing data; performing multimodal fusion and data compression on the multimodal sensing data to determine lightweight monitoring data; transferring the lightweight monitoring data to the lightweight data model to evaluate and determine the sleep state, wherein the sleep state is identified by a collection timestamp.

[0029] The multimodal sensing data undergoes multimodal fusion and lightweight data compression, and is then transferred to the lightweight data model for sleep state assessment. This process includes: performing sensor acquisition in the operating mode of a set of peripheral devices to obtain time-series acquisition data from each sensor in the operating mode, thus determining the multimodal sensing data. Subsequently, multimodal fusion is performed on the multimodal sensing data to align the sensor data in time. After processing, data compression is performed to determine lightweight monitoring data. The lightweight monitoring data is then transferred to the lightweight data model, and the model output is obtained to determine the sleep state, wherein the sleep state is identified by a data acquisition timestamp.

[0030] The method provided in this application embodiment further includes: setting a compressed frame rate; performing frame rate extraction and static feature identification on the modal fusion data based on the compressed frame rate to determine single-point data; identifying dynamic features between frame rates to determine data activity; integrating the single-point data and the data activity in positive time sequence to obtain the lightweight monitoring data; performing analog-to-digital conversion on the lightweight monitoring data and transmitting it to the lightweight data model.

[0031] The goal of data compression is to optimize data transmission and storage, enabling the system to effectively monitor sleep states with low power consumption while maintaining high processing speed and accuracy. Setting a compression frame rate, i.e., the frequency of data acquisition, such as performing frame rate extraction every three time intervals to acquire one set of data. In sleep monitoring, setting the compression frame rate controls the amount of data collected. Setting an appropriate compression frame rate is crucial for reducing power consumption and data volume; specific compression frame rate settings can be configured by professional technicians. Subsequently, based on the compression frame rate, modal fusion data is extracted according to the compression frame rate. Simultaneously, static feature identification is performed on the modal fusion data, extracting multiple effective features determined from the spatial dimension, such as extracting the maximum, minimum, and mean values ​​of the data sequence, resulting in single-point data. This single-point data includes the frame rate extracted data and static features. Then, dynamic features between frame rates are identified to determine data activity. Dynamic features are the trend characteristics between two data groups. For example, dynamic features are obtained by interpolating adjacent data to determine the amplitude of data changes, identifying the changing characteristics, and thus determining data activity. The changes in each data point are recorded during the data activity. Furthermore, the single-point data and the data activities are integrated in chronological order to obtain the lightweight monitoring data. Finally, the lightweight monitoring data is converted from analog to digital, and then transmitted to the lightweight data model.

[0032] The method provided in this application embodiment further includes: drawing a sleep map with time phase and sleep state as the axis; identifying the acquisition timestamp of the output sleep state and updating the sleep map; generating sleep warning information based on the abnormal state determination of the sleep state and the abnormal trend determination of the state of the sleep map; and the wearable device responding to the sleep warning information by issuing an abnormal alarm.

[0033] The time-series integration output sleep map includes: The sleep map visualizes sleep states by plotting the trend of sleep states over time. Time-series integration combines the sleep state at each moment with time information to generate a complete sleep map for analyzing sleep quality. When plotting the sleep map, the time phase refers to different time periods during sleep. Each moment's sleep state can be identified using timestamps. Each sleep state corresponds to a time period. Through time-series integration, the sleep state of each time period is plotted into a map, with the horizontal axis representing time (phase) and the vertical axis representing sleep state. For example, between 12:00 AM and 12:30 AM, the user may be in a light sleep stage, while between 12:30 AM and 1:00 AM, they may be in a deep sleep stage. These states will be clearly displayed in the sleep map. Subsequently, the output sleep states are collected and timestamps are identified to obtain the timestamps corresponding to the output sleep states, and the sleep map is updated based on the timestamps. Finally, anomaly detection is performed based on the sleep states to determine whether there are any abnormalities in the duration of the sleep state or the proportion of time spent in each phase. The system performs anomaly detection based on the changes in the sleep spectrum, judging the trend of sleep state changes and checking whether the sleep state exhibits irregular changes. For example, a user should enter light sleep after deep sleep, but if the system detects that the user suddenly jumps from deep sleep to wakefulness, it may indicate a sleep quality problem. Based on the anomaly detection results, a sleep warning message is generated, and the wearable device responds to the sleep warning message by issuing an anomaly alert.

[0034] The method provided in this application embodiment further includes: introducing new monitoring nodes based on the sleep map, wherein the new monitoring nodes are sampling time nodes and node sensor sources inserted based on the monitoring rules, and the node sensor source is at least one of the sensor sources. Based on the new monitoring nodes, user sleep monitoring and management are performed.

[0035] After outputting the sleep map, the process includes: if the current data fluctuates significantly, or if a certain physiological characteristic shows some abnormality but does not meet the criteria for an abnormal state, then monitoring nodes are added to enable targeted data monitoring, facilitating subsequent analysis of the abnormality. Based on the sleep map, new monitoring nodes are introduced. These new monitoring nodes are sampling time nodes and node sensor sources inserted on top of the monitoring rules; that is, at least one node sensor source is added to the monitoring rules. Finally, based on the newly added monitoring nodes, user sleep monitoring and management are performed.

[0036] In the preceding text, a method for processing sleep data of a wearable device supporting low power consumption according to an embodiment of the present invention was described in detail with reference to FIG1. ​​Next, a system for processing sleep data of a wearable device supporting low power consumption according to an embodiment of the present invention will be described with reference to FIG2.

[0037] According to an embodiment of the present invention, a sleep data processing system for wearable devices supporting low power consumption solves the technical problem that existing sleep monitoring devices have high power consumption and are difficult to accurately monitor the user's sleep state under low power conditions. By dynamically dividing sleep stages, allocating different sensors and sampling frequencies, and combining lightweight data models and multimodal fusion data compression technology, efficient sleep state monitoring under low power conditions is achieved, while improving the system's response speed and monitoring accuracy. The low-power wearable device sleep data processing system includes: a stage division module 11, a model building module 12, a sensor data acquisition module 13, and a state evaluation module 14.

[0038] The stage division module 11 is used to divide sleep stages, allocate monitoring sampling frequencies and sensor sources to each sleep stage, and set monitoring rules. The sleep stages are divided based on sleep depth.

[0039] The model building module 12 is used to build a low-power module using a sample-driven training method. The low-power module is built into the processor of the wearable device and includes a front-end data interface and a lightweight data model.

[0040] The sensor data acquisition module 13 is used to perform mode management on the processor's peripheral devices based on the monitoring rules, and to monitor and acquire multimodal sensor data.

[0041] The state assessment module 14 is used to perform multimodal fusion and lightweight data compression on the multimodal sensing data based on the front-end data interface, and transfer it to the lightweight data model for sleep state assessment, and output a sleep map by time-series integration, wherein the compression method is to extract single-point data and data activity.

[0042] The specific configuration of the sensor data acquisition module 13 will be described in detail below. The sensor data acquisition module 13 may further include: managing the mode of the processor's peripheral devices based on the monitoring rules, including: introducing a power management module, which performs mode switching of the processor's peripheral devices, including operating mode and standby mode; identifying the monitoring rules and writing I / O control commands, wherein the I / O control commands identify the device code and mode type; and managing the mode of the peripheral devices based on the I / O control commands, wherein the mode management includes one set of operating modes for peripheral devices and two sets of standby modes for peripheral devices.

[0043] The specific configuration of the model building module 12 will be described in detail below. The model building module 12 further includes: a low-power module comprising a lightweight data model, including: calling sleep state evaluation records and integrating to determine training samples, wherein the training samples include input-multi-level intermediate evaluation-output. Based on the training samples, supervised training of the initialization data model is performed. The initialization data model is subjected to lightweight processing and validated based on the training samples to determine the lightweight data model, wherein the lightweight processing method is the reduction of the number of network layers and parameter types, and a reduction constraint is imposed with a preset influence degree.

[0044] The specific configuration of the primary verification module 14 will be described in detail below. The primary verification module 14 may further include: performing multimodal fusion and lightweight data compression on the multimodal sensing data, and transferring it to the lightweight data model for sleep state evaluation, including: performing sensor acquisition in the working mode of a set of peripheral devices to determine the multimodal sensing data; performing multimodal fusion and data compression on the multimodal sensing data to determine lightweight monitoring data; transferring the lightweight monitoring data to the lightweight data model to evaluate and determine the sleep state, wherein the sleep state is identified by a data acquisition timestamp.

[0045] The specific configuration of the first-level verification module 14 will be described in detail below. The first-level verification module 14 further includes: performing data compression to determine lightweight monitoring data, including: setting a compression frame rate; based on the compressed frame rate, performing frame rate extraction and static feature identification on the modal fusion data to determine single-point data; identifying dynamic features between frame rates to determine data activity; integrating the single-point data and the data activity in positive time sequence to obtain the lightweight monitoring data; and performing analog-to-digital conversion on the lightweight monitoring data and transmitting it to the lightweight data model.

[0046] The specific configuration of the first-level verification module 14 will be described in detail below. The first-level verification module 14 further includes: the time-series integrated output sleep map, which includes: drawing a sleep map with phase and sleep state as axes; identifying the collection timestamps of the output sleep states and updating the sleep map; generating sleep warning information based on the abnormal state determination of the sleep states and the abnormal trend determination of the sleep map state; and the wearable device responding to the sleep warning information by issuing an anomaly alarm.

[0047] The specific configuration of the first-level verification module 14 will be described in detail below. The first-level verification module 14 further includes: after outputting the sleep map, it further includes: based on the sleep map, introducing new monitoring nodes, wherein the new monitoring nodes are sampling time nodes and node sensor sources inserted based on the monitoring rules, and the node sensor source is at least one of the sensor sources. Based on the new monitoring nodes, user sleep monitoring and management are performed.

[0048] The wearable device sleep data processing system provided in this embodiment of the invention can execute the wearable device sleep data processing method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0049] While this application makes various references to certain modules in the system according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved. In addition, the specific names of each functional unit are only for easy distinction and are not intended to limit the scope of protection of this invention.

[0050] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for processing sleep data in wearable devices that supports low power consumption, characterized in that, The method includes: dividing sleep stages, allocating monitoring sampling frequencies and sensor sources to each sleep stage, and setting monitoring rules, wherein sleep stages are divided based on sleep depth; constructing a low-power module using sample-driven training, the low-power module being built into the processor of the wearable device, including a front-end data interface and a lightweight data model; performing mode management on the processor's peripheral devices based on the monitoring rules, monitoring and acquiring multimodal sensor data; performing multimodal fusion and lightweight data compression on the multimodal sensor data based on the front-end data interface, transferring the data to the lightweight data model for sleep state assessment, and time-series integrated output of a sleep map, wherein compression is performed by extracting single-point data and data activity.

2. The method for processing sleep data of wearable devices supporting low power consumption as described in claim 1, characterized in that, Based on the monitoring rules, mode management is performed on the processor's peripheral devices, including: introducing a power management module, which performs mode switching of the processor's peripheral devices, including working mode and standby mode; identifying the monitoring rules and writing I / O control commands, wherein the I / O control commands identify the device code and mode type; and performing mode management on the peripheral devices based on the I / O control commands, wherein mode management includes a set of working modes for peripheral devices and two sets of standby modes for peripheral devices.

3. The method for processing sleep data of wearable devices supporting low power consumption as described in claim 1, characterized in that, The low-power module includes a lightweight data model, comprising: calling sleep state assessment records, integrating and determining training samples, wherein the training samples include input-multi-level intermediate assessment-output; supervising the training of an initial data model based on the training samples; performing lightweight processing on the initial data model, and validating it based on the training samples to determine the lightweight data model, wherein the lightweight processing method is the reduction of the number of network layers and parameter types, and the reduction is constrained by a preset influence.

4. The method for processing sleep data of wearable devices supporting low power consumption as described in claim 1, characterized in that, The process of performing multimodal fusion and lightweight data compression on the multimodal sensing data, and then transferring it to the lightweight data model for sleep state assessment, includes: performing sensing acquisition in the working mode of a set of peripheral devices to determine the multimodal sensing data; performing multimodal fusion and data compression on the multimodal sensing data to determine lightweight monitoring data; and transferring the lightweight monitoring data to the lightweight data model to assess and determine the sleep state, wherein the sleep state is identified by a data acquisition timestamp.

5. A method for processing sleep data in a wearable device that supports low power consumption, as described in claim 4, characterized in that... Data compression is performed to determine lightweight monitoring data, including: setting a compression frame rate; based on the compression frame rate, performing frame rate extraction and static feature identification on the modal fusion data to determine single-point data; identifying dynamic features between frame rates to determine data activity; integrating the single-point data and the data activity in positive time sequence to obtain the lightweight monitoring data; performing analog-to-digital conversion on the lightweight monitoring data and transmitting it to the lightweight data model.

6. The method for processing sleep data of wearable devices supporting low power consumption as described in claim 5, characterized in that, The time-series integrated output sleep map includes: drawing a sleep map with phase and sleep state as the axis; identifying the collection timestamp of the output sleep state and updating the sleep map; generating sleep warning information based on the abnormal state determination of the sleep state and the abnormal trend determination of the state change of the sleep map; and the wearable device responding to the sleep warning information by issuing an abnormal alarm.

7. The method for processing sleep data of wearable devices supporting low power consumption as described in claim 1, characterized in that, After outputting the sleep map, the process includes: introducing new monitoring nodes based on the sleep map, wherein the new monitoring nodes are sampling time nodes and node sensor sources inserted based on the monitoring rules, and the node sensor source is at least one of the sensor sources; and performing sleep monitoring and management for users based on the new monitoring nodes.

8. A sleep data processing system for wearable devices that supports low power consumption, characterized in that, The system includes: a stage segmentation module, used to divide sleep stages, allocate monitoring sampling frequencies and sensor sources to each sleep stage, and set monitoring rules, wherein sleep stages are divided based on sleep depth; a model building module, used to build a low-power module using sample-driven training, wherein the low-power module is built into the processor of the wearable device, including a front-end data interface and a lightweight data model; a sensor data acquisition module, used to manage the mode of the processor's peripheral devices based on the monitoring rules, monitor and acquire multimodal sensor data; and a state evaluation module, used to perform multimodal fusion and lightweight data compression on the multimodal sensor data based on the front-end data interface, transfer the data to the lightweight data model for sleep state evaluation, and output a sleep map through time-series integration, wherein compression is performed by extracting single-point data and data activity.

Images