Methods and systems for improving the measurement of sleep data by classifying users based on sleeper type

Abstract

The present disclosure relates to systems and methods for improving analysis of sleep data by classifying users based on sleeper type. In particular, a wearable computing system can obtain a first set of motion sensor data for a user during a first time period from a motion sensor. The wearable computing system can determine a sleeper type for the user from a plurality of sleeper types based on the first set of motion sensor data received from the motion sensor. The wearable computing system can select a sleep analysis model from a plurality of sleep analysis models based on the sleeper type determined for the user. The wearable computing system can analyze a second set of motion sensor data from a second time period using the selected sleep analysis model to determine one or more sleep features for the user during the second time period.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to improving the accuracy of data analysis of data collected by sensors in a wearable computing device. More specifically, the present disclosure relates to methods and systems for improving the measurement of sleep data by classifying a user based on a sleeper type. BACKGROUND

[0002] Recent advances in wearable technology, such as fitness bands and smartwatches, have enabled the collection of data from a user based on sensors included in a wearable computing device. One use of the collected data can be to analyze a user’s sleep patterns. For example, using data collected by motion sensors, such as accelerometers, a wearable computing device can estimate the amount and quality of sleep of a user during a particular sleep session. However, a user’s sleep patterns and behaviors can vary, so systems and tools that accurately determine sleep data for some types of users can not be accurate when used to analyze sleep data generated by another type of user. SUMMARY

[0003] Aspects and advantages of embodiments of the present disclosure will be set forth in part in the following description, or can be learned from the description, or can be learned through practice of the embodiments.

[0004] One example embodiment includes a wearable computing device. The wearable computing device includes a motion sensor configured to generate motion sensor data based on movement of a user. The wearable computing device includes one or more processors that execute computer-readable instructions to obtain, from the motion sensor, a first set of motion sensor data for the user during a first time period. The one or more processors can further execute the computer-readable instructions to determine, from the first set of motion sensor data received from the motion sensor, a sleeper type for the user from a plurality of sleeper types. The one or more processors can further execute the computer-readable instructions to select, based on the sleeper type determined for the user, a sleep analysis model from a plurality of sleep analysis models. The one or more processors can further execute the computer-readable instructions to analyze, using the selected sleep analysis model, a second set of motion sensor data from a second time period to determine one or more sleep characteristics of the user during the second time period.

[0005] Another example embodiment can include a computer-implemented method comprising obtaining, by a computing device having one or more processors, a first set of motion sensor data for a user during a first time period from a motion sensor. The method further includes determining, by the computing device, a sleeper type for the user from a plurality of sleeper types based on the first set of motion sensor data received from the motion sensor. The method further includes selecting, by the computing device, a sleep analysis model from a plurality of sleep analysis models based on the sleeper type determined for the user. The method further includes analyzing, by the computing device, a second set of motion sensor data from a second time period using the selected sleep analysis model to determine one or more sleep characteristics for the user during the second time period.

[0006] Another example embodiment can be a non-transitory computer-readable storage medium having computer-readable program instructions embodied thereon that when executed by one or more processors cause the one or more processors to obtain, from a motion sensor, a first set of motion sensor data for a user during a first time period. The instructions can further cause the one or more processors to determine a sleeper type for the user from a plurality of sleeper types based on the first set of motion sensor data received from the motion sensor. The instructions can further cause the one or more processors to select a sleep analysis model from a plurality of sleep analysis models based on the sleeper type determined for the user. The instructions can further cause the one or more processors to analyze a second set of motion sensor data from a second time period using the selected sleep analysis model to determine one or more sleep characteristics for the user during the second time period.

[0007] Other example aspects of the present disclosure relate to systems, apparatuses, computer program products (such as tangible, non-transitory computer-readable media, but also such as software downloadable over a communications network without having to be stored in non-transitory form), user interfaces, memory devices, and electronic devices for determining a sleeper type of a user.

[0008] These and other features, aspects, and advantages of various embodiments will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the related principles. BRIEF DESCRIPTION OF DRAWINGS

[0009] A detailed discussion of embodiments of the application in relation to the drawings is presented in the following description of the embodiments. The discussion and description relate to only some embodiments of the application and are not intended to exclude other embodiments or applications of the application.

[0010] Figure 1 FIGURE 1 illustrates an example wearable computing device in accordance with example embodiments of the present disclosure.

[0011] Figure 2The illustration shows a block diagram of an example computing environment including a wearable computing device with sensors, according to an example embodiment of the present disclosure.

[0012] Figure 3 The diagram illustrates a block diagram of a computing environment including a wearable computing device connected to a server computing system via a network, according to an example embodiment of the present disclosure.

[0013] Figure 4 This is a block diagram of an example process for determining a user's sleeper type according to an example embodiment of the present disclosure.

[0014] Figure 5 The illustration shows example time-series motion data according to an example embodiment of the present disclosure.

[0015] Figure 6 The illustration shows a representation of a possible sleep state according to an example embodiment of the present disclosure.

[0016] Figure 7 An example process for determining a user's sleeper type according to an example embodiment of the present disclosure is illustrated.

[0017] Figure 8 A block diagram depicts an example gesture classification machine learning model according to an example embodiment of the present disclosure.

[0018] Figure 9 This is a flowchart depicting an example process for determining a user's sleeper type based on motion data according to an example embodiment of the present disclosure. Detailed Implementation

[0019] Reference will now be made in detail to embodiments, one or more examples of which are illustrated in the accompanying drawings. Each example is provided as an explanation of the embodiments and not as a limitation thereof. In fact, those skilled in the art will understand that various modifications and variations can be made to the embodiments without departing from the scope or spirit of this disclosure. For example, features illustrated or described as part of one embodiment may be used with another embodiment to produce further embodiments. Thus, various aspects of this disclosure are intended to cover such modifications and variations.

[0020] An exemplary aspect of this disclosure relates to a system for improving the classification of sleep data by: determining a user's sleeper type; and selecting a sleep analysis model based on the user's sleeper type to more accurately interpret the sleep data (e.g., motion sensor data). For example, some users move significantly more during sleep (e.g., highly mobile or twitching sleepers) than is typically expected of other users. As a result, sleep data generated by the sleep analysis model may be inaccurate for such highly mobile users. Therefore, to improve accuracy, the user classification system can determine the user's sleeper type and select a trained or otherwise designed sleep analysis model for the user's specific sleeper type to accurately determine sleep characteristics (e.g., information about the user's sleep segments or sleep habits).

[0021] Wearable computing devices may include motion sensors (e.g., accelerometers) that detect movement by the user wearing the device. The wearable computing device may collect motion sensor data from the user during one or more sleep segments. Based on this data (e.g., past sleep segments), the computing device may determine the user's sleeper type. For example, the computing device may determine the average amount of time between detected movements by the user during a sleep segment. If the average amount of time between movements is below a threshold, the user can be identified as having a high-movement sleeper type (e.g., a twitching sleeper type). Once a particular user's sleeper type is determined, the user classification system can select a specific sleep analysis model from multiple sleep analysis models for use when analyzing future sleep data received from that user.

[0022] For example, a user can wear a wearable computing device (e.g., a fitness tracker) to detect metrics that can represent information about the amount and quality of sleep in relation to the user experience. To accurately determine one or more sleep characteristics based on sensor data generated by the wearable computing device, a user classification system (e.g., included in the wearable computing device itself or at a remote computing system) can use past sleep data to determine the user's sleeper type. For example, the user classification system can determine the user's sleeper type based on one or more sleep factors. One or more sleep factors can be determined based on previously collected sleep data from the user. In some examples, one or more sleep factors may include the average amount of time between movements, the minimum or average number of movements the user makes within any 100-minute time interval of a sleep segment, and / or other indicators of the amount, frequency, and / or duration of movements during sleep.

[0023] A user classification system can compare one or more sleep factors with one or more high-mobility thresholds. If one or more sleep factors exceed the threshold, the user classification system can determine that the user's sleeper type is a high-mobility sleeper. In response, the user classification system can choose a sleep analysis model already configured for high-mobility sleeper users or another analysis tool.

[0024] Once a sleep analysis model has been selected, the user classification system can obtain further motion sensor data from the user and use this data as input to the selected sleep analysis model to determine sleep characteristic data. For example, the user classification system can determine one or more sleep characteristics describing the user's sleep segments at the time the further motion sensor data was obtained, including whether the user is asleep, the user's current sleep state, whether the user has transitioned between sleep states, and if so, when those transitions occurred.

[0025] More specifically, wearable computing devices can include any computing device integrated into an object intended to be worn by a user. For example, wearable computing devices can include, but are not limited to, smartwatches, fitness trackers, computing devices integrated into jewelry such as smart rings or smart necklaces, computing devices integrated into clothing items such as jackets, shoes, and trousers, and wearable glasses with computing elements included therein. In some examples, wearable computing devices may include one or more sensors intended to collect information with the permission of the user wearing the wearable computing device.

[0026] In some examples, the wearable computing device may include one or more motion sensors. The motion sensors may include accelerometers that can measure the user's movement along three axes (e.g., the x, y, and z dimensions). In some examples, the wearable computing device may transmit motion sensor data to a server system for analysis. In other examples, the analysis may be performed by the wearable computing device itself. Therefore, the user classification system may be enabled or executed at one or more locations within the wearable computing device, a remote server system, or another computing system that can perform analysis on the motion sensor data and communicate with the wearable computing device to receive the motion sensor data.

[0027] User classification systems included in computing systems (e.g., remote server systems or wearable computing devices) can determine the range of movement a user makes in each of the three dimensions over a specific time period (e.g., 30-second intervals). This range of movement at specific intervals can be used to determine a sleep coefficient. The range can be determined by identifying a minimum value for each axis during the specific time period, identifying a maximum value for each axis, and subtracting the minimum value from the maximum value.

[0028] The sleep coefficient can be determined based on the user's movement during a time interval. For example, it can be generated by summing the range of movement across all three axes. In another example, the sleep coefficient can be set as the maximum range of movement across all three axes. In yet another example, the sleep coefficient can be set as the average range of movement across all three axes.

[0029] In some examples, other sensors can be used to detect a user's movement. For instance, infrared sensors can be used to sense a user's movement. Additionally, other factors, such as a user's breathing, can be measured by infrared sensors and can be used to estimate the user's movement and otherwise determine the user's sleep data. Similarly, pedometers can be used to determine a user's sleep state (e.g., a walking user is generally not asleep). In some examples, sensors can measure a user's heart rate and use that data to determine the user's movement. Similarly, sensors included in wearable computing devices can be used to perform photoplethysmography (PPG). PPG data can be used to determine the user's movement.

[0030] Based on a sleep coefficient, a user classification system can determine whether a given interval includes the user's movement. In some examples, the system can determine the amount of movement a user makes during a time period that includes several intervals. For instance, if each interval is 30 seconds long, the system can determine whether the user fell asleep within a 10-minute time period based on the multiple intervals included in that period. In some examples, if the amount of movement exceeds a certain threshold, the system can determine that the user did not fall asleep.

[0031] In some examples, sleep coefficients for a given time interval can be sorted into one of several categories or bins, each bin representing a predetermined range of sleep coefficient values. Each bin can be associated with a specific determination about the user. Thus, the smallest bin can be associated with no movement during the interval and can be associated with sleep with a high degree of confidence. Similarly, sleep coefficients above a certain level can be categorized.

[0032] A user classification system can generate time series of motion sensor data representing multiple sequential intervals, where each interval is classified as including or excluding movement. This time series of motion sensor data representing multiple sequential intervals can be referred to as a sleep log. The user classification system can determine a user's sleeper type based on the time series of motion sensor data. Therefore, to determine a user's sleeper type, the user classification system can obtain motion sensor data for sleep segments. The user classification system can determine one or more factors based on the motion sensor data for sleep segments. Factors may include the average amount of time between movements during a sleep segment, the minimum or average sleep coefficient over any 100-minute interval during the sleep segment, and / or other indicators of the amount, frequency, or duration of movement during sleep.

[0033] Based on one or more sleep factors (e.g., the minimum number of movements during a time period and the average amount of time between movements), a user classification system can determine a user's sleeper type. For example, if a user's number of movements during a sleep period exceeds a certain threshold, the user classification system can determine that the user's sleeper type is a high-movement sleeper (e.g., a twitching sleeper). Other potential sleeper types may be associated with sleep disorders (e.g., insomnia), may be based on the user's sleep habits (e.g., a user with several short sleep segments rather than one long sleep segment), and so on.

[0034] Once a sleeper type is determined for a user, the user classification system can select a specific sleep analysis model to use when analyzing the user's sleep data. For example, the user classification system may include multiple different computer-learned models, each trained based on data from or associated with a specific sleeper type. Therefore, when determining a user's sleeper type, the user classification system can select a trained sleep analysis model to more accurately analyze motion sensor data or other sleep data generated by users with that sleeper type.

[0035] In other examples, a single sleep analysis model is used, but the values ​​of specific parameters associated with the model can be adjusted based on the user's sleeper type. Therefore, when determining a user's sleeper type, the user classification system can select specific parameters associated with that sleeper type and use those parameters during the analysis of sleep data associated with that user.

[0036] Alternatively, in some examples, the sleep analysis model may include (or be used as a post-processing system) one or more rule-based classification steps that can perform additional analyses on the sleep characteristic data determined by the sleep analysis model under one or more specific conditions. In some examples, the rule-based classification steps may be modified based on the user's sleeper type (e.g., such that they produce different sleep characteristics for users of a first sleeper type and users of a second sleeper type). In this way, the user classification system can select the most appropriate tool to analyze the user's motion sensor data based on the user's sleeper type.

[0037] Using motion sensor data detected from the user during a second time period as input to a sleep analysis model (e.g., data from a first time period used to determine the user's sleeper type), a user classification system can determine one or more sleep characteristics for the second time period. In some examples, the user classification system can determine whether the user fell asleep during one or more segments of the second time period. If so, the user classification system can determine when the sleep segment began and ended. Additionally, the user classification system can determine the user's current sleep state. Sleep states can include REM sleep, light sleep, and deep sleep. In some examples, other sleep characteristics can be determined based on the sleep analysis model. For example, the sleep analysis model can generate sleep scores or other data representing the overall quality of the sleep segments (e.g., based on the length of the sleep segment and the time spent in each sleep state). Additional parameters not particularly relevant to sleep can be used as factors in their analysis using data from the sleep analysis model, such as stress scores and / or fitness readiness scores.

[0038] The embodiments of the disclosed technology provide numerous technical effects and benefits, particularly in the field of user computing devices. In particular, the embodiments of the disclosed technology provide improved techniques for analyzing sensor data associated with a sleeping user. For example, using the embodiments of the disclosed technology, a computing system can determine the user's sleeper type and, based on the sleeper type, select a sleep analysis model to use when analyzing motion sensor data generated by a wearable computing device worn by the user. Customizing the sleep analysis model to evaluate data from a specific user can increase the accuracy of sleep characteristics generated by the computing system and can result in a better and more useful user experience. Furthermore, this effect is achieved at a relatively low cost. Therefore, the disclosed embodiments enable additional functionality to be implemented without significantly increasing the overall cost of the wearable device.

[0039] Exemplary aspects of this disclosure will now be discussed in more detail with reference to the accompanying drawings.

[0040] Figure 1A front view of an example wearable computing device 100 according to an example embodiment of the present disclosure is depicted. In one embodiment, the wearable computing device 100 may be a wristband, bracelet, watch, armband, ring placed around a user's finger, or other wearable product that may be equipped with sensors as described in the present disclosure. In an example embodiment, the wearable computing device 100 is configured with a display 102, a device housing 104, a band 106, and one or more sensors. In an embodiment, the display 102 may be configured to present data to a user relating to the user's skin temperature, heart rate, sleep status, electroencephalogram (EEG), electrocardiogram (ECG), electromyogram (EMG), electrooculogram (EOG), and other physiological data of the user (e.g., blood oxygen level). The display 102 may also be configured to convey data from additional environmental sensors included within the wearable computing device 100. Example information from these additional environmental sensors conveyed on the display 102 may include location, altitude, and weather associated with the user's position. The display 102 may also convey data about the user's movement (e.g., whether the user is stationary, walking, and / or running).

[0041] In an example embodiment, display 102 can be configured to receive data input by a user. In this embodiment, a user can request the wearable computing device 100 to generate additional data (e.g., sleep data) for display to the user via input on the display. In response, the display can present an instruction to the user (e.g., an instruction to place one or more fingers on sensor 310) to obtain the requested data. Furthermore, if the wearable computing device 100 determines that additional data is necessary while it is being collected, display 102 can present an instruction to the user (e.g., displaying "Please continue to place your fingers on the sensor for 10 seconds").

[0042] In an example embodiment, the device housing 104 may be configured to include one or more sensors described in this disclosure. Example sensors included by the device housing 104 may include motion sensors (e.g., accelerometers), pulse oximeters, IR motion sensors, skin temperature sensors, internal device temperature sensors, location sensors (e.g., GPS), altitude sensors, heart rate sensors, audio sensors, pressure sensors, gyroscopes, environmental sensors (e.g., bedside ultrasound sensors), and other physiological sensors (e.g., blood oxygen level sensors). In embodiments, the device housing 104 may also be configured to include one or more processors. The strap 106 may be configured to secure the wearable computing device 100 around a user's arm by connecting the two ends of the strap 106, for example, by means of a buckle, clip, or other similar fastening device, thereby allowing the wearable computing device 100 to be worn by the user.

[0043] Figure 2The illustration depicts an example computing environment including a wearable computing device 100 according to an exemplary embodiment of the present disclosure. In this example, the wearable computing device 100 may include one or more processors 202, a memory 204, one or more sensors 210, a user classification system 212, a sleep analysis system 220, and a display system 222.

[0044] More specifically, one or more processors 202 can be any suitable processing device that can be embedded within the form factor of the wearable computing device 100. For example, such processors 202 can include one or more processor cores, microprocessors, application-specific integrated circuits (ASICs), FPGAs, controllers, microcontrollers, etc. One or more processors 202 can be a single processor or multiple processors operatively connected. Memory 204 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, etc., and combinations thereof.

[0045] Specifically, in some devices, memory 204 may store instructions for implementing user classification system 212, sleep analysis system 220, and / or display system 222. Thus, wearable computing device 100 may implement user classification system 212, sleep analysis system 220, or display system 222 to perform various aspects of this disclosure.

[0046] It should be understood that the term "system" can refer to specialized hardware, computer logic executing on a more general-purpose processor, or a combination thereof. Thus, a system can be implemented in hardware, application-specific integrated circuits (ASICs), firmware, and / or software that controls a general-purpose processor. In one embodiment, the system can be implemented as a program code file stored on a storage device, loaded into memory, and executed by a processor, or it can be provided from a computer program product (such as computer-executable instructions) stored in a tangible computer-readable storage medium (such as RAM, a hard disk, or optical or magnetic media).

[0047] The memory 204 may also include data 206 and instructions 208 that can be retrieved, manipulated, created, or stored by one or more processors 202. In some example embodiments, such data may be accessed and used as input to the user classification system 212, the sleep analysis system 220, and / or the display system 222. In some examples, the memory 204 may include data for performing one or more processes, as well as instructions describing how those processes can be performed.

[0048] In some examples, wearable computing device 100 may include one or more sensors 210. For example, sensor 210 may include, but is not limited to, one or more: motion sensors (e.g., accelerometers), pulse oximeters, IR motion sensors, skin temperature sensors, internal device temperature sensors, location sensors (e.g., GPS), altitude sensors, heart rate sensors, audio sensors, pressure sensors, and other physiological sensors (e.g., blood oxygen level sensors).

[0049] In some examples, one or more sensors 210 can detect a user's movement and provide data representing that movement to the wearable computing device 100. For example, an accelerometer can measure the user's movement across three axes. For example, an accelerometer can measure the X, Y, and Z axes. Thus, as the user moves, the accelerometer can generate data representing that movement and provide it to the wearable computing device 100. Other sensors, such as IR laser sensors, heart rate sensors, pulse oximeters, etc., can be used to determine the user's movement or activity level.

[0050] Motion data generated by sensor 210 can be transmitted to user classification system 212. User classification system 212 can use the recorded motion data to estimate the user's sleeper type. For example, the recorded motion data can be longitudinal data previously recorded from the user and identified as being associated with the user's sleep segments. Specifically, motion data processing system 214 can use the motion data to generate sleep coefficients for specific time periods. In some examples, the specific time period can be a 30-second time window. Other time increments can be used. Motion data processing system 214 can determine the amount of movement within a corresponding time interval based on motion data generated by one or more sensors 210.

[0051] For example, motion data processing system 214 can calculate the total movement in each axis that occurs during a corresponding time interval. The sleep coefficient can represent the sum of movements on all three axes during the corresponding time interval. For example, motion data processing system 214 can determine the minimum and maximum values ​​for each axis during the corresponding time interval. The total range of movement along each axis can be determined by subtracting the minimum value from the maximum value of the axis.

[0052] Therefore, the motion data processing system 214 can determine the user's total range of motion during that time increment. In some examples, the total range for each axis can be summed to generate a sleep coefficient. Alternatively or additionally, the sleep coefficient can be the average of the total range of motion detected for all three axes. In yet another example, the sleep coefficient can be set to the highest of three different range values ​​from the three axes. Alternatively or additionally, the sleep coefficient can also be based at least in part on data other than the total range of motion. For example, one or more sensors can provide data describing whether the wearable computing device is currently being worn by the user (e.g., whether a fitness band is currently on the user's wrist). Additionally, one or more sensors can determine the number of steps taken by the user during that time interval. These additional data inputs can be used to generate or update the motion coefficient for a specific time interval.

[0053] Once the sleep coefficients for the corresponding intervals have been determined, the motion data processing system 214 can classify the specific interval as including or excluding movement. In some examples, the sleep coefficients can be sorted into one of multiple bins of sleep coefficient values, each bin representing a range of sleep coefficient values ​​and associated with a specific amount of movement. Thus, similar sleep coefficient values ​​can be grouped together, and movement can be determined based on rules defined for each bin.

[0054] Grouping or representing sleep coefficients into bins can reduce the amount of processing time and power required to generate sleep factor information without significantly sacrificing accuracy. The range represented by each bin can vary, such that for small coefficient ranges (e.g., representing small movements), the range of each bin can be relatively small. In contrast, for high movement ranges (e.g., when the user is making significant movements), the range of the bins can be larger, because at high movement levels the user is almost certainly awake.

[0055] The motion data processing system 214 can generate a series of sleep coefficients, which can be referred to as a sleep log, for a series of consecutive time intervals. For example, the motion data processing system 214 can generate sleep coefficients for a sequential time interval of 40 minutes and store them as a sleep log. Other time lengths can be used. Each time interval in the sequence can be specified as including or excluding movement based on its associated sleep coefficient (or based on bins sorted by its associated sleep coefficients). The data in the sleep log can be used as input to a machine learning model to determine whether a user fell asleep during the time period represented by the sleep log. However, if the machine learning model has not been trained for a specific sleeper type, the model's output may be inaccurate. Therefore, the user classification system 212 can determine the user's sleeper type to ensure that their sleep log data is analyzed accurately.

[0056] The time interval sequence can be transmitted to the sleeper type determination system 218. The sleeper type determination system 218 can determine the user's sleeper type based on the classification of each time interval in the time interval sequence of motion sensor data. The sleeper type determination system 218 can determine one or more factors based on motion sensor data from sleep segments. These factors may include the average amount of time between movements during a sleep segment, the minimum or average number of movements within any 100-minute time interval during a sleep segment, and / or other indications of the amount, frequency, and / or duration of movements during sleep.

[0057] Based on one or more sleep factors, the sleeper type determination system 218 can determine a user's sleeper type. For example, if the number of instances of user movement during a sleep period exceeds a certain threshold, the sleeper type determination system 218 can determine that the user's sleeper type is a high-movement sleeper type (e.g., a twitching sleeper type). Other potential sleeper types may be associated with sleep disorders (e.g., insomnia), may be based on the user's sleep habits (e.g., a user with several short sleep segments rather than one long sleep segment), and so on. It should be noted that the sleeper type determination system 218 is described as primarily using previously collected sleep movement data to determine a user's sleeper type. However, other data, such as demographic data, location data, ambient temperature data, etc., may also be used as part of this determination.

[0058] User classification system 212 can transmit the determined sleeper type to sleep analysis system 220. In some examples, user classification system 212 can store the determined sleeper type in memory 204, and sleep analysis system 220 can access memory 204 as needed.

[0059] The sleep analysis system 220 can select a specific sleep analysis model to use when analyzing a user's sleep data. For example, the sleep analysis system 220 can access multiple different computer learning models, each of which has been trained based on data from or associated with a specific sleeper type. Thus, when determining a user's sleeper type, the sleep analysis system 220 can select a trained sleep analysis model to analyze motion sensor data or other sleep data generated by a user with that sleeper type during a sleep segment.

[0060] Alternatively, a single sleep analysis model may be used, but the specific parameter values ​​associated with the model can be adjusted based on the user's sleeper type. Thus, when the user's sleeper type is determined, the sleep analysis system 220 can select specific parameters associated with the sleeper type and set the model parameters to those specific parameters during the analysis of sleep data associated with the user.

[0061] Alternatively, in some examples, the sleep analysis model may include (or be used as a post-processing system) one or more rule-based classification steps that can perform additional analysis on the sleep feature data generated by the sleep analysis model under one or more specific conditions. In some examples, the rule-based classification steps may be modified based on the user's sleeper type (e.g., such that they produce different sleep features for users of a first sleeper type and users of a second sleeper type). In this way, the sleep analysis system 220 can select the most appropriate tool to analyze the user's motion sensor data based on the user's sleeper type.

[0062] Examples of rule-based classification steps could include post-processing rules that change a user's classification from asleep to awake for a given time interval if the sleep coefficient for that interval exceeds a predetermined threshold. However, if a user is identified as a high-mobility user, the predetermined threshold value can be increased or the entire post-processing rule can be disabled. In another example, a post-processing rule could change the classification of a sleep period from asleep to awake if a certain percentage of included sleep intervals have a sleep coefficient above a threshold. Similar to the first example, if a user is identified as a high-mobility user, the threshold for the rule-based classification step can be increased or the rule itself can be paused. Therefore, in addition to selecting a specific sleep analysis model based on a user's sleeper type, hard-coded post-processing rules can also be modified (e.g., threshold adjusted) or paused so that they are not applied to high-mobility sleeper type users.

[0063] Using motion sensor data detected from the user during a second time period as input to the sleep analysis system 220 (e.g., data from a first time period used to determine the user's sleeper type), the sleep analysis system 220 can determine one or more sleep characteristics during the second time period. For example, the first set of sleep data may be the user's past sleep data (e.g., data from multiple sleep segments), while the second set of sleep data may be data generated by sensors during the user's most recent sleep segment.

[0064] In some examples, sleep analysis system 220 can determine whether a user falls asleep during one or more segments of a second time period. If so, sleep analysis system 220 can determine when the sleep segment begins and ends. Additionally, sleep analysis system 220 can determine one or more sleep states the user experiences during the sleep segment, the amount of time spent in each sleep state, and the time it takes for the user to transition from one sleep state to another. Sleep states can include REM sleep, light sleep, and deep sleep. In some examples, other sleep characteristics can be determined based on a sleep analysis model. For example, the sleep analysis model can generate a sleep score representing the overall quality of the sleep segment (e.g., based on the length of the sleep segment and the time spent in each sleep state). Additional parameters not particularly relevant to sleep can be used as a factor in their analysis using data from the sleep analysis model, such as stress scores and / or fitness readiness scores.

[0065] Once one or more sleep data have been generated, the sleep analysis system 220 can provide one or more sleep parameters to the display system 222. The display system 222 can present one or more sleep parameters to a user via a display included in the wearable computing device 100. In some examples, a user can request the wearable computing device 100 to display one or more sleep parameters via an interface with the device. The wearable computing device 100 can present sleep parameters to the user (e.g., sleep volume, sleep quality, duration of sleep segments, etc.).

[0066] Figure 3 An example client-server environment according to an exemplary embodiment of this disclosure is depicted. The client-server system environment 300 includes one or more wearable computing devices 100 and a server computing system 330. One or more communication networks 320 may interconnect these components. The communication network 320 may be any of a variety of network types, including a local area network (LAN), a wide area network (WAN), a wireless network, a wired network, the Internet, a personal area network (PAN), or a combination of such networks.

[0067] Wearable computing device 100 may include, but is not limited to, smartwatches, fitness trackers, computing devices integrated into jewelry such as smart rings or smart necklaces, computing devices integrated into clothing items such as jackets, shoes, and trousers, and wearable glasses with computing elements included therein. In some examples, wearable computing device 100 may include one or more sensors that are intended to collect information with the permission of a user wearing the wearable computing device. In some examples, wearable computing device 100 may connect to another computing device, such as a personal computer (PC), laptop computer, smartphone, tablet computer, mobile phone, electrical components of a vehicle, or any other electronic device capable of communicating with communication network 320. Wearable computing device 100 may include one or more user applications, such as search applications, communication applications, navigation applications, productivity applications, gaming applications, word processing applications, or any other applications. User applications may include web browsers. Wearable computing device 100 may use a web browser (or other applications) to send requests to and receive requests from server computing system 330.

[0068] In some examples, wearable computing device 100 may include one or more sensors that can be used to determine user movement at a specific time. For example, sensor 210 may include, but is not limited to: motion sensors (e.g., accelerometers), pulse oximeters, IR motion sensors, skin temperature sensors, internal device temperature sensors, location sensors (e.g., GPS), altitude sensors, heart rate sensors, audio sensors, pressure sensors, and other physiological sensors.

[0069] like Figure 3 As shown, the server computing system 330 can typically be based on a three-tier architecture, consisting of a front-end layer, an application logic layer, and a data layer. Those skilled in the art in the relevant computer and internet fields should understand that... Figure 3 Each component shown can represent a set of executable software instructions and the corresponding hardware (e.g., memory and processor) for executing those instructions. To avoid unnecessary detail, [details omitted]. Figure 3 Various components and engines irrelevant to conveying an understanding of the various examples are omitted. However, those skilled in the art will readily recognize that various additional components and engines can be used with the server computing system 330, such as Figure 3 As shown, this is to facilitate additional functionality not specifically described herein. Furthermore, Figure 3 The various components described can reside on a single server computer, or they can be distributed across several server computers in various arrangements. Furthermore, although the server computing system 330... Figure 3 The architecture is described as having a three-layer structure, but various example implementations are by no means limited to this architecture.

[0070] like Figure 3 As shown, the front end may consist of an interface system 322 that receives communications from one or more wearable computing devices 100 and conveys appropriate responses to them. For example, the interface system 322 may receive Hypertext Transfer Protocol (HTTP) requests or other web-based, application programming interface (API) requests. The wearable computing device 100 may execute regular web browser applications or applications already developed for a specific platform, encompassing any of a wide variety of mobile devices and operating systems.

[0071] like Figure 3 As shown, the data layer may include a sleep analysis database 134. In some example embodiments, the sleep analysis database 134 may store various types of data, including but not limited to data associated with multiple sleep analysis models (e.g., one or more sleep analysis models for each sleeper type). The sleep analysis database 134 may include sleep and motion data received from the wearable computing device 100 and representing data collected by one or more sensors 210 included in the wearable computing device 100. The sleep analysis database 134 may also include data generated by the user classification system 212, such as movement classification for one or more time intervals (e.g., classification of whether a particular time interval includes user movement). The sleep analysis database 134 may also include information generated by the sleep analysis system 220, including determination of whether the user fell asleep at a particular time, the length and quality of the user's sleep, etc.

[0072] The application logic layer may include application data, which can provide a wide range of other applications and services, allowing users to access or receive geographic data for navigation or other purposes. The application logic layer may include a user classification system 212 and a sleep analysis system 220.

[0073] User classification system 212 can receive motion data from sensors 210 included in wearable computing device 100 via information interface system 322. User classification system 212 can analyze the motion data to determine whether the user moved during each of a plurality of time intervals. In some examples, user classification system 212 can generate one or more sleep factors based on the motion data and information about whether the user moved in each of the plurality of time intervals. Based on one or more sleep factors, user classification system 212 can identify the sleeper type associated with the user.

[0074] The sleep analysis system 220 can be used to select a specific sleep analysis model to use when analyzing a user's sleep data once the user's sleeper type has been identified. For example, the sleep analysis system 220 can access multiple different computer-learned models (e.g., in the sleep analysis database 134), each of which has been trained based on data from or associated with a specific sleeper type. Thus, when a user's sleeper type is determined, the sleep analysis system 220 can select a trained sleep analysis model to more accurately analyze motion sensor data or other sleep data generated by a user with that sleeper type during a sleep period.

[0075] Once a sleep analysis model has been selected, the sleep analysis system 220 can receive further motion data from sensors 210 included in the wearable computing device 100. Using this detected motion sensor data as input to the sleep analysis model (e.g., data from a first time period used to determine the user's sleeper type), the sleep analysis system 220 can determine one or more sleep characteristics for a second time period. In some examples, the sleep analysis system 220 can determine whether the user fell asleep during the second time period. If so, the sleep analysis system 220 can determine when the sleep segment began and when it ended.

[0076] Additionally, the sleep analysis system 220 can determine the user's current sleep state. Sleep states can include REM sleep, light sleep, and deep sleep. In some examples, other sleep characteristics can be determined based on a sleep analysis model. For instance, the sleep analysis model can generate a sleep score representing the overall quality of sleep segments (e.g., based on the length of the sleep segment and the time spent in each sleep state). Additional characteristics not particularly relevant to sleep, such as stress scores and / or fitness readiness scores, can be determined. One or more sleep characteristics can be transmitted to the wearable computing device 100 for display to the user.

[0077] By sending sensor data captured at a specific wearable computing device 100 to a server system for analysis, the wearable computing device 100 can minimize the amount of memory and power used at its location while still receiving sleep characteristics determined by machine learning models using significant resources. Centralizing the analysis of sleep data to a server system improves the user experience while minimizing the cost of the wearable computing device 100.

[0078] Figure 4 This is a block diagram 400 of an example process for determining a user's sleeper type according to an example embodiment of the present disclosure. In this example, a user classification system (e.g., Figure 2The user classification system 212) can obtain previously detected motion data of the user at 402. For example, the user classification system can access the user's stored historical motion data, including at least one sleep segment or suspected sleep segment.

[0079] Previously detected motion data may include data from motion sensors (e.g., Figure 2 The motion data is generated by sensor 210. In some examples, the motion sensor is an accelerometer, and the motion data is represented as the range of motion along three axes (x, y, and z axes) during the time period in which the data was acquired. For example, the accelerometer can measure acceleration in three dimensions and correlate the measured acceleration with a specific timestamp (or another method of determining when a particular motion occurred). In some examples, user classification system 212 can divide the motion data into one or more time intervals. For example, each time interval can represent 30 seconds. Other time lengths can be used for the time intervals. User classification system 212 can generate a classification for each corresponding time interval, indicating whether the user was detected as moving during that corresponding time interval.

[0080] Once the user's motion data has been analyzed, the user classification system 212 can determine the user's sleeper type at point 404 based on the analysis of the motion data. In some examples, the sleeper type can be a high-mobility sleeper type. In a specific example of a high-mobility sleeper type, determination can be based on the number of times the user moves within a specific interval (e.g., within one hour). In a specific example of a high-mobility sleeper type, the number of times the user moves within a specific interval can be compared to a threshold. If the number of moves during the corresponding interval exceeds the threshold, the user classification system 212 can determine that the user is a high-mobility sleeper type user.

[0081] Once a sleeper type has been determined for a user, the sleep analysis system 220 can select a sleep analysis model based on that sleeper type at point 406. Therefore, the sleep analysis model used to analyze the motion data of a user with a high-mobility sleeper type at point 408 will differ from the sleep analysis model selected for a user who is not a high-mobility sleeper type. This allows the sleep analysis model to accurately determine the user's sleep characteristics, which would otherwise be difficult to determine based on the user's sleeper type. For example, if a sleep analysis model trained for users with sleeper types other than high-mobility sleeper type is used to analyze the motion data of a high-mobility user, the sleep analysis model may not be able to detect the high-mobility user's sleep at all during the entire sleep period.

[0082] The sleep analysis system 220 can use a selected sleep analysis model to determine one or more sleep characteristics of a user. For example, sleep characteristics may include determining whether the user is asleep at a specific time, the start and end of a sleep segment, the specific sleep state the user is in, and so on.

[0083] Figure 5 An example of time-series motion data according to an exemplary embodiment of the present disclosure is illustrated. In this example, the entire sleep segment 506 can be divided into a plurality of time intervals 502. Each corresponding interval of the plurality of time intervals 502 can be classified based on a sleep coefficient (based on sensor data captured during the time period represented by the corresponding time interval). The sleep coefficient can be generated based on the range of motion detected by the accelerometer during the time interval 502. If the range of motion exceeds a threshold, the time interval 502 can be associated with the user's movement. If the range of motion does not exceed the threshold, the time interval 502 can be associated with no movement.

[0084] In some examples, one or more time intervals 502 may be grouped into sleep periods 504. A sleep period 504 may refer to a set of time intervals 502 that together comprise a duration less than the entire sleep segment 506 but more than a single time interval 502. In some examples, a sleep period 504 may represent a sleep segment within a sleep segment 506 separated by one or more shift time intervals.

[0085] Figure 6 The illustration depicts a representation of a possible sleep state according to an example embodiment of the present disclosure. For example, one possible state is a waking state 602 in which the user is determined to be awake. In some examples, the waking state 602 may be identified based on multiple activity metrics including user interaction with a wearable computing device or based on the amount of movement detected from the user exceeding a predetermined threshold for movement.

[0086] In some examples, sleep state 610 may include one or more different states, such as REM state 612, characterized by rapid eye movement and low muscle tone in the body (e.g., residual muscle tone). Another sleep state may be light sleep state 614, which may be characterized by muscle relaxation, cessation of eye movement, decreased respiratory and heart rates, decreased body temperature, and slowed brain waves. During light sleep state 614, the user may be relatively easily awakened.

[0087] The third sleep state can be deep sleep state 616. Deep sleep state 616 can be characterized by further relaxation of muscles, a lower respiratory rate and heart rate than in light sleep state, a further slowing of brain waves, and difficulty in waking the user from sleep.

[0088] Figure 7The illustration depicts an example process for determining a user's sleeper type according to an example embodiment of this disclosure. In some examples, a user classification system (e.g., Figure 2 The user classification system 212 can obtain the user's motion data during a time period at point 702. In some examples, the time period can be the period during which the user classification system 212 determines that the user has fallen asleep.

[0089] User classification system 212 can analyze motion data at 704 points to determine one or more sleep factors. For example, motion data can be analyzed to determine when the user fell asleep and when the user was awake during the motion data period. User classification system 212 can determine the duration of each movement by the user during the sleep period. As described above, user movement can be determined based on a variety of different sensors, including accelerometers that measure user movement, infrared sensing systems that detect user movement in space, heart monitors that determine user activity based on user heartbeats, and any other sensors capable of estimating user movement based on signals obtained from the user.

[0090] In some examples, a sleeper type determination system can determine one or more sleep factors at point 704 based on motion data. A sleep factor can be any measure of the duration of a user's sleep quality that is useful in determining a user's sleeper type. Specific examples may include the number of times a user moves during a one-hour sleep period. In some examples, the number of times a user moves per hour may be averaged over the entire sleep period. Another factor may include the average length of time between movements. Besides the two specific examples given here, a variety of other measurements can be used as sleep factors when assessing a user's sleeper type. For example, any data that indicates the amount, frequency, or duration of movement during sleep can be used as a potential sleep factor.

[0091] In some examples, one or more sleep factors can be used as input to a classification system that can determine a user's sleeper type based on multiple inputs. One or more sleep factors can be used directly or to generate sleep factor scores based on motion data. User classification system 212 can determine at 706 whether one or more sleep factors (or sleep factor scores) exceed a threshold. The threshold can be a predetermined value used to distinguish one sleeper type from another. If one or more sleep factors (or sleep factor scores) exceed the threshold, the user classification system can determine at 708 that the user's sleeper type is a high-mobility sleeper type. Based on this determination, user classification system 212 can select a sleep analysis model associated with the high-mobility sleeper type user at 710.

[0092] Based on the determination that the sleep factor (or sleep score) does not exceed a threshold, the user classification system 212 can determine at point 712 that the user is not a high-mobility sleeper. The user classification system 212 can then select a sleep analysis model associated with users who were not identified as high-mobility sleepers at point 714.

[0093] Once a sleep analysis model has been selected, the sleep analysis system 220 can use the selected sleep analysis model to determine sleep parameters at 716 based on motion data captured by one or more sensors.

[0094] Figure 8 A block diagram depicts an example sleep analysis model according to an exemplary embodiment of the present disclosure. The machine learning sleep analysis model 800 can obtain motion data from a user as input data 806. For example, the sleep analysis model 800 can generate one or more sleep parameters based on the input motion data. Once trained, the machine learning sleep analysis model can output 808 one or more sleep factors, such as a determination of whether the user is asleep during a specific time period represented by the input motion data.

[0095] In some examples, the machine learning data analysis model 1810 may otherwise include various machine learning models, such as neural networks (e.g., deep neural networks), other types of machine learning models, including nonlinear and / or linear models, or binary classifiers. Neural networks may include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or other forms of neural networks.

[0096] Various training techniques can be used to train the machine learning sleep analysis model 800. Specifically, the machine learning sleep analysis model 800 can be trained using one of several semi-supervised training techniques. The machine learning sleep analysis model 800 can also be trained using supervised training techniques, such as error backpropagation. For example, a loss function can be backpropagated through the model to update one or more parameters of the model (e.g., based on the gradient of the loss function). Various loss functions can be used, such as mean squared error, likelihood loss, cross-entropy loss, hinge loss, and / or various other loss functions. Gradient descent techniques can be used to iteratively update the parameters over several training iterations. In some implementations, performing error backpropagation can include performing truncated backpropagation over time. Generalization techniques (e.g., weight decay, dropouts, etc.) can be performed to improve the generalization ability of the model being trained.

[0097] Figure 9This is a flowchart depicting an example process for determining a user's sleeper type based on motion data according to an example embodiment of this disclosure. In some examples, the wearable computing device (e.g., a fitness band or smartwatch) 100 may include a motion sensor configured to generate motion sensor data based on the user's movement. The motion sensor may be an accelerometer. An accelerometer may measure movement (or acceleration) along each of three axes. The wearable computing device may include a user classification system (e.g., Figure 2 User classification system 212) and sleep analysis system (e.g., Figure 2 The sleep analysis system 220 in the example 212 can obtain a set of motion sensor data of a user from the motion sensor at point 902 during a first time period. In some examples, the first time period includes one or more sleep segments.

[0098] User classification system 212 can determine sleep coefficient data based on the range of motion detected by an accelerometer during one or more time intervals. User classification system 212 can classify users as either moving or stationary during each corresponding time interval of one or more time intervals.

[0099] User classification system 212 can determine a user's sleeper type from multiple sleeper types at point 904 based on motion sensor data received from a motion sensor. In some examples, the multiple sleeper types may include a high-mobility sleeper type. User classification system 212 can determine a user's sleeper type based on motion sensor data received from a motion sensor by determining the average number of user movements during a first time period. User classification system 212 can determine that the average number of user movements during the first time period exceeds a high-mobility threshold. In response to determining that the average number of user movements during the first time period exceeds the high-mobility threshold, user classification system 212 can determine that the user's sleeper type is a high-mobility sleeper type.

[0100] User classification system 212 can determine a user's sleeper type based on motion sensor data received from a motion sensor by determining the average amount of time between time intervals in which the user is classified as mobile (one or more time intervals). User classification system 212 can determine that the average amount of time between time intervals in which the user is classified as mobile exceeds a high-mobility threshold. In response to determining that the average amount of time between time intervals in which the user is classified as mobile exceeds the high-mobility threshold, user classification system 212 can determine that the user's sleeper type is a high-mobility sleeper type.

[0101] The user classification system 212 can select a sleep analysis model from multiple sleep analysis models at point 906 based on the sleeper type determined for the user. In some examples, the multiple sleep analysis models can be machine learning models. In some examples, each of the multiple sleep analysis models is trained to accurately analyze data from different sleeper types among multiple sleeper types. In some examples, the sleep analysis model includes one or more post-processing rules.

[0102] In some examples, the sleep analysis system 220 may, at point 908, analyze a second set of motion sensor data from a second time period using a selected sleep analysis model to determine one or more sleep characteristics of the user during the second time period. In some examples, the one or more sleep characteristics include data indicating whether the user was asleep during the second time period. In some examples, the one or more sleep characteristics include data indicating the user's sleep state during the second time period. In some examples, the one or more sleep characteristics include data indicating the length of a sleep segment.

[0103] The technologies discussed herein relate to servers, databases, software applications, and other computer-based systems, as well as the actions taken and the information sent to and received from such systems. Those skilled in the art will recognize that the inherent flexibility of computer-based systems allows for a wide variety of possible configurations, combinations, and divisions of tasks and functions between and within components. For example, the server processes discussed herein can be implemented using a single server or multiple servers working in combination. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0104] While the subject matter has been described in detail with respect to specific exemplary embodiments thereof, it should be understood that modifications, variations, and equivalents of such embodiments can be readily derived by those skilled in the art upon acquiring an understanding of the foregoing. Therefore, the scope of this disclosure is illustrative rather than limiting, and the subject matter of this disclosure does not exclude such modifications, variations, and / or additions to the subject matter, which will be apparent to those skilled in the art.

Claims

1. A wearable computing device, comprising: A motion sensor configured to generate motion sensor data based on the user's movement; as well as One or more processors, the one or more processors executing computer-readable instructions to: The first set of motion sensor data of the user during the first time period is obtained from the motion sensor; Based on the first set of motion sensor data received from the motion sensor, the sleeper type is determined for the user from multiple sleeper types; A sleep analysis model is selected from multiple sleep analysis models based on the sleeper type determined for the user, wherein the multiple sleep analysis models are machine learning models, and each of the multiple sleep analysis models is trained to analyze data from different sleeper types among the multiple sleeper types; as well as The selected sleep analysis model is used to analyze a second set of motion sensor data from a second time period to determine one or more sleep characteristics of the user during the second time period.

2. The wearable computing device according to claim 1, wherein, The first time period includes one or more sleep segments.

3. The wearable computing device according to claim 1, wherein, The motion sensor is an accelerometer.

4. The wearable computing device according to claim 3, wherein, The one or more processors determine the sleeper type for the user from multiple sleeper types by the following steps based on the first set of motion sensor data received from the motion sensor: generating sleep coefficient data based on the range of motion detected by the accelerometer during the first set of motion sensor data.

5. The wearable computing device according to claim 4, wherein, The processor: The user is classified as mobile or stationary during each of one or more time intervals.

6. The wearable computing device according to claim 5, wherein, The multiple sleeper types include the high-mobility sleeper type.

7. The wearable computing device according to claim 6, wherein, Based on the first set of motion sensor data received from the motion sensors, the one or more processors determine the sleeper type for the user through the following steps: Determine the average number of user movements during the first time period; as well as Determine that the average number of user moves during the first time period exceeds a high movement threshold; as well as In response to determining that the average number of user movements during the first time period exceeds a high movement threshold, the user's sleeper type is determined to be the high-movement sleeper type.

8. The wearable computing device according to claim 6, wherein, Based on the first set of motion sensor data received from the motion sensors, the one or more processors determine the sleeper type for the user through the following steps: Determine the average time interval between which the user is classified as mobile; and It was determined that the average amount of time between the time intervals in which the user was classified as mobile was below a high mobility threshold; as well as In response to determining that the average amount of time between time intervals in which the user was classified as mobile is below a high mobility threshold, the user's sleeper type is determined to be the high mobility sleeper type.

9. The wearable computing device according to any one of claims 1 to 8, wherein, The selected sleep analysis model includes one or more post-processing rules.

10. The wearable computing device according to any one of claims 1 to 8, wherein, The one or more sleep features include data indicating whether the user fell asleep during the second time period.

11. The wearable computing device according to any one of claims 1 to 8, wherein, The one or more sleep features include data indicating the user's sleep state during the second time period.

12. The wearable computing device according to any one of claims 1 to 8, wherein, The one or more sleep characteristics include data indicating the length of sleep segments.

13. A computer-implemented method, the method comprising: The first set of motion sensor data for the user during a first time period is obtained from the motion sensor using a computing device including one or more processors. Using the computing device, based on the first set of motion sensor data received from the motion sensor, a sleeper type is determined for the user from multiple sleeper types; Using the computing device, a sleep analysis model is selected from multiple sleep analysis models based on a sleeper type determined for the user, wherein the multiple sleep analysis models are machine learning models, and each of the multiple sleep analysis models is trained to analyze data from different sleeper types among the multiple sleeper types; as well as The computing device uses a selected sleep analysis model to analyze a second set of motion sensor data from a second time period to determine one or more sleep characteristics of the user during the second time period.

14. The computer-implemented method according to claim 13, wherein, The first time period includes one or more sleep segments.

15. The computer-implemented method according to any one of claims 13 and 14, wherein, The motion sensor is an accelerometer.

16. The computer-implemented method according to claim 15, further comprising: The computing device generates sleep coefficient data based on the range of motion detected by the accelerometer during one or more time intervals.

17. A non-transitory computer-readable storage medium having computer-readable program instructions embodied thereon, the computer-readable program instructions causing the one or more processors, when executed, to: Obtain the first set of motion sensor data from the motion sensor during the first time period; Based on the first set of motion sensor data received from the motion sensor, the sleeper type is determined for the user from multiple sleeper types; A sleep analysis model is selected from multiple sleep analysis models based on the sleeper type determined for the user, wherein the multiple sleep analysis models are machine learning models, and each of the multiple sleep analysis models is trained to analyze data from different sleeper types among the multiple sleeper types; as well as The selected sleep analysis model is used to analyze a second set of motion sensor data from a second time period to determine one or more sleep characteristics of the user during the second time period.

18. The non-transitory computer-readable storage medium according to claim 17, wherein, The multiple sleeper types include the high-mobility sleeper type.

Images