Motion metric generation in wearable electronics

Abstract

This application relates to movement metric generation in wearable electronic devices. Aspects of generating movement metrics quantifying motion data samples generated by a wearable electronic device are discussed herein. For example, in one aspect, an embodiment can obtain motion data samples generated by a set of motion sensors in a wearable electronic device. The wearable electronic device can then generate a first movement metric based on a quantification of a first set of motion data samples. The wearable electronic device can also generate a second movement metric based on a quantification of a second set of motion data samples. The wearable electronic device can then cause a non-transitory machine-readable storage medium to store the first movement metric and the second movement metric as time series data.

Description

[0001] This application is a divisional application of Chinese Patent Application No. 201910277916.4, filed on September 23, 2015, entitled "Generation of Motion Measurement in Wearable Electronic Device", which is a divisional application of Chinese Patent Application No. 201510614306.0, filed on September 23, 2015, entitled "Wearable Electronic Device, Method Performed Thereon and Computer-Readable Storage Device".

[0002] Cross-reference of related applications

[0003] This application claims the benefit of the following provisional applications: U.S. Provisional Application No. 62 / 054,380, filed September 23, 2014, which is incorporated herein by reference in its entirety; U.S. Provisional Application No. 62 / 063,941, filed October 14, 2014, which is incorporated herein by reference in its entirety; U.S. Provisional Application No. 62 / 067,914, filed October 23, 2014, which is incorporated herein by reference in its entirety; and U.S. Provisional Application No. 62 / 068,622, filed October 24, 2014, which is incorporated herein by reference in its entirety. Technical Field

[0004] The embodiments relate to the field of wearable electronic devices. Specifically, the embodiments relate to automated motion detection using wearable electronic devices. Background Technology

[0005] Wearable electronic devices have become increasingly popular among consumers. These devices use various sensors to track user activity and help users maintain a healthy lifestyle. To determine user activity, wearable electronic devices collect activity data and perform calculations on that data. One challenge in accurately determining user activity is that these wearable electronic devices, because they are worn by the user, are typically packaged in compact housings containing processors that are less powerful than those in larger electronic devices (on which complex calculations are more difficult to perform).

[0006] Many wearable electronic devices track metrics about specific activities, such as step counts for running and walking. Other metrics that can be tracked by wearable electronic devices include metrics about sleep. Typically, for tracking sleep initiation metrics, wearable electronic devices may include an interface that provides the user with notifications of when they plan to transition to a sleep state (e.g., via the user pushing a button on the wearable electronic device or tapping on the wearable electronic device). Summary of the Invention

[0007] One aspect of the present invention is to provide a wearable electronic device to be worn by a user. The wearable electronic device includes: a group of one or more motion sensors for generating motion data samples representing motion of the wearable electronic device, the motion data samples comprising a first set of one or more motion data samples generated during a first time interval and a second set of one or more motion data samples generated during a second time interval; a group of one or more processors coupled to the group of motion sensors; and a non-transitory machine-readable storage medium coupled to the group of one or more processors and storing instructions therein, which, when executed by the group of one or more processors, cause the group of one or more processors to: obtain the motion data samples generated by the group of motion sensors; generate a first motion measure based on quantization of the first set of motion data samples; generate a second motion measure based on quantization of the second set of motion data samples; and cause the non-transitory machine-readable storage medium to store the first motion measure and the second motion measure as time-series data.

[0008] Preferably, when the instruction is executed by the group one or more processors, the group one or more processors cause the group one or more processors to: assign a sleep state to the user of the wearable electronic device for a time block covering the first time interval based on an analysis including the first motion metric and the second motion metric.

[0009] Preferably, when the instructions are executed by the group one or more processors, the group one or more processors perform the analysis including the first motion metric and the second motion metric by causing the group one or more processors to perform the following operations: deriving values ​​of features corresponding to a time period, the values ​​being calculated from motion metrics within a time window including the first motion metric and the second motion metric; and selecting, based on the analysis of the values ​​of the features for a plurality of time periods covered by the time block, the plurality of time periods including the time period.

[0010] Preferably, when the instruction is executed by the group one or more processors, it also causes the group one or more processors to perform the analysis of the value of the feature for the plurality of time periods covered by the time block by causing the group one or more processors to perform the following operation: determining the level of the user's activity for the plurality of time periods, wherein the determination of the activity level for an individual time period is based on a comparison of the value of the feature corresponding to the individual time period with a threshold, and the level is used in the selection of the sleep state.

[0011] Preferably, the quantization of the first set of motion data samples includes transforming the first set of motion data samples into a single numerical value.

[0012] Preferably, the quantification of the motion data includes statistical measures of the motion data along the motion axis.

[0013] Preferably, the statistical measure is one or more of the following: quantile, interquartile range, measure of deviation, and measure of entropy.

[0014] Preferably, the statistical measure is derived from the frequency domain analysis of the motion data.

[0015] Preferably, the frequency domain analysis quantifies the amount of periodic motion and compares the amount of periodic motion in two frequency bands.

[0016] Preferably, the quantization of the motion data includes a function of the maximum value and the minimum value of the motion data along the axis.

[0017] Preferably, the function of the maximum motion data and the minimum motion data along the axis includes the application of a weighted value.

[0018] Preferably, the quantization of the motion data includes the calculation of the standard deviation of the motion data along the axis.

[0019] Preferably, the quantization of the motion data includes the calculation of the time derivative of the motion data along the axis.

[0020] Preferably, the quantization of the motion data includes a combination of interquartile ranges of the motion data along the axis.

[0021] Another aspect of the present invention is to provide a method performed by a wearable electronic device. The method includes: obtaining a first set of one or more motion data samples generated by a group of one or more motion sensors, the first set of motion data samples being generated during a first time interval; generating a first motion metric based on quantization of the first set of motion data samples; obtaining a second set of one or more motion data samples generated by the group of motion sensors, the second set of motion data samples being generated during a second time interval; generating a second motion metric based on quantization of the second set of motion data samples; and causing a non-transitory machine-readable storage medium to store the first motion metric and the second motion metric as time-series data.

[0022] Preferably, the method further includes: assigning sleep states to the user of the wearable electronic device for time blocks covering the first time interval based on analysis including the first motion metric and the second motion metric.

[0023] Preferably, the analysis including the first movement metric and the second movement metric includes: deriving values ​​of a feature corresponding to a time period, the values ​​being calculated from movement metrics within a time window including the first movement metric and the second movement; and selecting, based on the analysis of the values ​​of the feature for multiple time periods covered by the time block, the multiple periods including the time period.

[0024] Preferably, the analysis of the values ​​of the features for the plurality of time periods covered by the time blocks includes: determining the level of activity of the user for the plurality of time periods, wherein the determination of the activity level for an individual time period is based on a comparison of the value of the feature corresponding to the individual time period with a threshold, and the level is used in the selection of the sleep state.

[0025] Preferably, quantization of the first set of motion data samples includes transforming the first set of motion data samples into a single numerical value.

[0026] Preferably, the quantification of the motion data includes statistical measures of the motion data along an axis.

[0027] Preferably, the statistical measure is one or more of the following: quantile, interquartile range, measure of deviation, and measure of entropy.

[0028] Preferably, the statistical measure is derived from the frequency domain analysis of the motion data.

[0029] Preferably, the frequency domain analysis quantifies the amount of periodic motion and compares the amount of periodic motion in two frequency bands.

[0030] Preferably, the quantization of the motion data includes a function of the maximum value and the minimum value of the motion data along the axis.

[0031] Another aspect of the present invention provides an electronic device comprising: a group of one or more processors; and a non-transitory machine-readable storage medium coupled to the group of one or more processors and wherein instructions are stored therein, which, when executed by the group of one or more processors, cause the group of one or more processors to: obtain a first motion metric, the first motion metric being a first set of motion data generated by a group of one or more motion sensors of a wearable electronic device worn by a user during a first time interval; obtain a second motion metric, the second motion metric being a second set of motion data generated by the group of motion sensors of the wearable electronic device worn by the user during a second time interval; classify the level of the user's activity for the first time interval based on an analysis of the first motion metric and the second motion metric; and determine a sleep state for a time block covering the first time interval using the activity level.

[0032] Preferably, when the instruction is executed, it causes the group one or more processors to analyze the first movement metric and the second movement metric based on the value of the derived statistical feature, the value of the statistical feature being calculated based on a statistical comparison of the values ​​of the first movement metric and the second movement metric, and the activity level being selected based on the value of the feature.

[0033] Preferably, when the instruction is executed, it causes the group one or more processors to quantize the first set of motion data based on summing the multidimensional acceleration data within the first time interval into a single value.

[0034] Preferably, the quantization of the first set of motion data includes transforming the first set of motion data into a single numerical value.

[0035] Preferably, the electronic device is external to and physically separate from the wearable electronic device, and the instruction, when executed, causes the group one or more processors to obtain the first motion measurement and the second motion measurement based on receiving the first motion measurement and the second motion measurement from the wearable electronic device via a wireless connection.

[0036] Preferably, the time block includes time intervals other than the first time interval, and the time intervals are classified using a mixture of different activity levels.

[0037] Preferably, the electronic device is the wearable electronic device. Attached Figure Description

[0038] The embodiments described herein are illustrated by way of example rather than by way of limitation in the accompanying drawings, in which the same references indicate similar elements.

[0039] Figure 1 This invention describes the use of motion metrics for user state detection and user sleep phase detection according to one embodiment of the invention.

[0040] Figure 2A This invention describes, according to one embodiment, the collection of motion metrics at different time windows for a time of interest, the determination of statistical characteristics of one of the times of interest, and the classification of the times of interest based on said statistical characteristics.

[0041] Figure 2B This invention describes how, according to one embodiment of the invention, user active / inactive states are classified based on mobility metrics, and how awake / sleeping states are further derived.

[0042] Figure 3A This describes a snapshot of a motion metric obtained from data over different time spans, according to an embodiment of the present invention.

[0043] Figure 3B This invention illustrates how data from different time spans can be used to determine statistical characteristics for a given moment of interest.

[0044] Figure 3C This invention describes how, according to one embodiment of the present invention, the active / inactive state of a user at a specific time point in a live system is classified during different time spans.

[0045] Figure 3D This invention illustrates how data from different time spans can be used to derive a user's awake / sleeping state according to one embodiment of the invention.

[0046] Figure 3E The invention describes an embodiment that includes a non-wearable state, which derives the user's active / inactive / non-wearable state at the live system during different time spans.

[0047] Figure 3F This invention illustrates that, according to one embodiment considering the non-wearable state, the user's awake / sleeping state is derived using data from different time spans.

[0048] Figure 4A According to one embodiment of the present invention, statistical characteristics are determined at different time windows for a time of interest, and statistical characteristics of one of the times of interest are determined and the times of interest are classified based on said statistical characteristics.

[0049] Figure 4B This describes the results of determining the sleep stage at different time windows according to an embodiment of the present invention.

[0050] Figure 5 A flowchart illustrating the generation of motion metrics and, optionally, user state detection and / or user sleep phase detection, according to an embodiment of the present invention.

[0051] Figure 6 A flowchart illustrating the automatic detection of a user's sleep cycle according to an embodiment of the present invention.

[0052] Figure 7 A flowchart illustrating the automatic detection of a user's sleep stage according to an embodiment of the present invention.

[0053] Figure 8A A flowchart illustrating the automatic reduction of power consumption of at least one of a photoplethysmography sensor and a motion sensor according to an embodiment of the present invention.

[0054] Figure 8B A flowchart illustrating the power consumption of at least one of a photoplethysmography sensor and a motion sensor according to an embodiment of the present invention.

[0055] Figure 9 A block diagram illustrating a wearable electronic device and an electronic device that implements the operations disclosed in an embodiment of the present invention.

[0056] Figure 10A This invention describes a wearable electronic device whose sides rest on a flat surface when not worn, according to one embodiment of the invention.

[0057] Figure 10B This describes a wearable electronic device according to an embodiment of the present invention, in which the front of the device is placed on a flat surface when not being worn.

[0058] Figure 10C This describes a wearable electronic device according to an embodiment of the present invention, which is placed on a flat surface such that it is perpendicular to the flat surface when not worn.

[0059] Figure 10D This invention describes a wearable electronic device whose back surface rests on a flat surface when not being worn, according to one embodiment of the invention.

[0060] Figure 10E This invention illustrates the orientation of a wearable electronic device when a user participates in various activities.

[0061] Figure 11 This is a flowchart illustrating an embodiment of the present invention, which uses an accelerometer to automatically detect when a wearable electronic device is not being worn.

[0062] Figure 12 Description for implementation from Figure 11 An exemplary alternative embodiment of frame 1104.

[0063] Figure 13 This describes an embodiment of the invention regarding the operation of non-wearable state detection using accelerated metrology.

[0064] Figure 14A According to one embodiment of the present invention, the time period is recorded as either a worn state or a non-wearable state based on acceleration data measured along the axis of the accelerometer exceeding a threshold acceleration for a continuous time period.

[0065] Figure 14B According to one embodiment of the present invention, the time span during which no wearable electronic device is worn is derived based on the state recorded for a continuous time period.

[0066] Figure 15A This invention describes the detection of a non-wearable state for a first axis time span according to an embodiment of the present invention.

[0067] Figure 15B This invention describes the detection of a non-wearable state for a second axis time span according to an embodiment of the present invention.

[0068] Figure 16 A block diagram illustrating a wearable electronic device and an electronic device that implements the operations disclosed in an embodiment of the present invention. Detailed Implementation

[0069] The embodiments described herein relate to the operation and features of wearable electronic devices. Wearable electronic devices can be configured to measure or otherwise detect movements experienced by the wearable electronic device. For simplicity of discussion, this movement can be referred to in the description of the display of the wearable electronic device, which is typically located on the user's forearm, in the same position that a wristwatch display would be positioned. While embodiments are described with reference to the fact that the display of the wearable electronic device is typically located on the user's forearm, in the same position that a wristwatch display would be positioned, the scope of the invention is not limited thereto, as modifications can be made to the wearable electronic device to make it wearable in different positions on the body (e.g., high on the forearm, on the opposite side of the forearm, on the leg, on the torso, as glasses, etc.), as will be understood by those skilled in the art.

[0070] Some exemplary embodiments may relate to a wearable electronic device that generates a motion metric based on motion data generated by the wearable electronic device's sensors. As used herein, "motion metric" can refer to data or logic that quantifies multiple samples of motion data generated by the wearable electronic device's motion sensors. In many cases, a motion metric will consume less memory to store compared to the sum of multiple samples of motion data quantified by the motion metric. In the case of accelerometer data, the motion metric can quantify acceleration data along one or more axes from multiple samples of accelerometer data. For example, the motion metric can be calculated every 30 seconds, where the motion metric quantifies samples of accelerometer data generated within those 30 seconds. Furthermore, the quantization can be a single number indicating the degree of motion experienced by the multiple samples along multiple axes.

[0071] While motion metrics can be used to reduce memory consumption from storing multiple samples of motion data, it's important to understand that motion metrics differ from simply compressing multiple samples of motion data. This is because compressed motion data typically requires decompression to become useful. That is, compressed data itself, without decompression, generally does not represent the degree of motion experienced during the time period in which the motion data was sampled. In contrast, the value of a motion metric itself can represent a concept of the degree of motion experienced during a timer interval corresponding to a sample of motion data. In this way, one motion metric can be compared with another to determine the difference in the degree of motion experienced during different time intervals.

[0072] In an exemplary embodiment, a wearable electronic device to be worn by a user may include a set of one or more motion sensors to generate motion data samples representing the movement of the wearable electronic device. The motion data samples include a first set of one or more motion data samples generated during a first time interval and a second set of one or more motion data samples generated during a second time interval. The wearable electronic device may be configured to (e.g., by executing instructions via a set of one or more processors) obtain the motion data samples generated by the set of motion sensors. The wearable electronic device may then generate a first motion measure based on the quantization of the first set of motion data samples. The wearable electronic device may also generate a second motion measure based on the quantization of the second set of motion data samples. The wearable electronic device may store the first and second motion measures as time-series data in a machine-readable storage medium.

[0073] Alternatively, some embodiments may automatically detect time blocks in which the wearer of the wearable electronics is asleep. Here, automatic detection may refer to detection occurring without explicit instructions from the wearer, such as navigating screen menus or otherwise entering the time period in which the wearer is sleeping. It should be understood that the term "automatic" does not exclude the possibility that the user may enable features that do not indicate a timing transition between different sleep states, or operate in a manner that does not indicate a timing transition between different sleep states. This embodiment of detecting time blocks in which the wearer is asleep may obtain a set of features for one or more time periods from motion data obtained from a set of one or more motion sensors, or data derived therefrom. Examples of data derived from motion data may include the movement metrics described above. The embodiment may then classify the one or more time periods into one of a plurality of user states based on the set of features determined for the one or more time periods. The state indicates the relative degree of movement of the user. In some cases, the state may be an enumerated state value, such as active or inactive, where each state represents a different degree of movement. In other cases, the state may be a numerical value or a range of values ​​that quantifies the degree of movement. The embodiments may then derive time blocks covering one or more time periods in which the user is in one of multiple states during this period. The states include awake and asleep states.

[0074] In some cases, certain embodiments are operable to adjust the power consumption of sensors based on the user's state. For example, based on the detection that the user's state, tracked by the wearable electronics, has changed to a sleeping state, one embodiment can reduce the power consumption of at least one sensor. Furthermore, based on the detection that the user's state has changed out of a sleeping state, one embodiment can reverse the reduction in the power consumption of the at least one sensor.

[0075] In some cases, certain embodiments are operable to detect when, in time, a wearable electronic device is not worn by a user. Some embodiments for detecting when a wearable electronic device is not worn may automatically determine the time period during which the wearable electronic device is not worn based on a comparison of motion data obtained from a set of motion sensors with a non-wear profile. The non-wear profile may be data or logic specifying a pattern of motion data indicating when the wearable electronic device is not worn by the user. These embodiments may then store data that associates the time period with the non-wearable state in a non-transitory machine-readable storage medium.

[0076] Please refer to the figures for examples and implementations.

[0077] Figure 1 This invention describes the use of motion metrics for user state detection and user sleep phase detection according to one embodiment of the invention. Figure 1Task frames and blocks can be implemented within the wearable electronic device or distributed between the wearable electronic device and one or more other electronic devices coupled to the wearable electronic device. The electronic device coupled to the wearable electronic device may be referred to herein as a secondary electronic device. A secondary electronic device may refer to a server (including hardware and software), a tablet computer, a smartphone (possibly running an application (called an app)), a desktop computer, or the like. The secondary electronic device may implement, for example, blocks 122 / 190 / 192. Alternatively, the secondary device providing sensor data may implement, for example, blocks 122 and / or 114. Task frames 1-5 illustrate operations that can be performed by... Figure 1 The order in which the components are executed is shown. However, it is understood that other embodiments may differ. Figure 1 The order in which tasks 1-5 are executed is shown in the image.

[0078] At task frame 1, a motion sensor (e.g., a multi-axis accelerometer) 112 generates motion data samples representing motion over multiple time intervals (e.g., the motion sensor 112 may be in a wearable electronic device and the motion data represents the motion of the device). The motion sensor may generate a number of motion data samples during a time interval. The number of samples generated in a time interval may depend on the sampling rate of the motion sensor and the length of the time interval. In the case of an accelerometer, the motion data samples may characterize a measure of acceleration along a movement axis. In some cases, the motion data samples may be data values ​​that consume a given amount of storage (e.g., 64-bit, 32-bit, 16-bit, etc.). The amount of storage consumed by the samples may depend on the accelerometer used in the embodiment and the implementation of the communication interface for transferring samples from the motion sensor to a microprocessor-accessible memory of the wearable electronic device.

[0079] At task frame 2, motion metric generator 116 generates multiple motion metrics based on motion data samples. Motion metrics can be used to reduce the storage space required to store data characterizing the movement of wearable electronics over time. For example, in one case, a motion metric can be quantified by a single value to represent the number of motion data samples generated for a given time period (e.g., within a 30-second time period). Furthermore, according to some embodiments, individual motion metrics may consume less storage space compared to a single motion data sample. For example, motion metrics may consume 4 bits, 8 bits, 16 bits, 32 bits, or any other storage size.

[0080] As discussed in more detail herein, several techniques exist for generating motion measurements. For example, in one embodiment, the motion measurement generator 116 may process motion data samples at the moment the sensor device generates samples, or alternatively, it may process blocks of motion data samples (e.g., samples for one or more time intervals) at a predetermined frequency or based on a number of samples.

[0081] The generation of motion metrics can be used to determine a user's sleep / wake state; therefore, motion metrics can also be referred to as sleep coefficients for determining sleep states in those embodiments. At least two basic versions of sleep detection methods exist. The first version is a motion sensor (e.g., accelerometer) solution that uses data from a triaxial motion sensor. The second version uses a combination of data from the motion sensor and an optical heart rate monitor (HRM) module including a photoplethysmography sensor 114.

[0082] At task box 3, the statistical measurer 122 can analyze motion metrics to derive a time series of values ​​characterizing one or more statistical features of a user's motion. In some cases, the feature value of a given motion metric may depend on motion metrics near the given motion metric (e.g., those that can be measured from a nearby location in the time series). For example, some embodiments of the statistical measurer 122 may utilize a scroll window of motion metrics to generate feature values ​​for a given motion metric. For illustration, assume that the set of motion metrics is represented as MM = {mm0, mm1, ..., mm}. n mm n+1 …mm z Therefore, the statistical measuring device 122 can measure the movement in mm. n-w-1 to mm n+w-1 Export mm n The eigenvalues ​​are given by a window size of 2*w. Examples of statistical features that can be derived from a statistical measuring instrument include (in particular) quantiles (e.g., quantiles of acceleration data), interquartile ranges, measures of deviation (e.g., Gini coefficient), measures of entropy, and information in the frequency domain (e.g., quantization of the amount of periodic motion and the amount of periodic motion in two or more frequency bands). Other examples of features are discussed below.

[0083] In optional task box 10, photoplethysmography sensor 114 can be used to generate photoplethysmography (PPG) data to calculate heart rate (HR), heart rate variability (HRV), and / or respiratory rate (RR). The photoplethysmography sensor typically includes a light source and a photodetector. The common light source is a light-emitting diode (LED).

[0084] At task frame 4A, once the values ​​of the group features have been derived, the user activity classifier 124 can classify or otherwise label the time periods corresponding to the statistical features with activity levels such as active or inactive. In some cases, the user activity classifier 124 classifies or otherwise labels each statistical feature in the group feature values; in other cases, the user activity classifier 124 can classify or otherwise label a subset of features in the time series of features (e.g., every other feature value, every five feature values, every ten feature values, etc.).

[0085] At task block 4B, time block classifier 126 assigns a dormant state selected from a plurality of sleep states to a time block using an activity level. Sleep states may include, in particular, awake and asleep states. As described in more detail below, some embodiments may include sleep states representing different sleep stages of a user. Furthermore, some embodiments may include sleep states representing different types of non-sleep activities (e.g., restlessness). Each time block may span one or more of the time periods characterized by an activity level. In some cases, the activity levels covered by the time block may be homogeneous activity levels, while in others heterogeneous activity levels are permitted.

[0086] therefore, Figure 1 This describes a system that can effectively and / or automatically determine when a wearer of a wearable electronic device is asleep. For example, instead of storing all motion data obtained from motion sensors and directly deriving sleep state from the raw motion data, the system can instead transform multiple motion data samples for a given time interval into a single value represented by a motion metric. Furthermore, in some cases, the activity level within a time period can be determined from multiple features derived from motion metrics across multiple time intervals.

[0087] Now described in more detail Figure 1 Operations and components.

[0088] Motion measurement

[0089] As described above, embodiments may use wearable electronics containing motion sensors to measure motion data (e.g., acceleration data measured by an accelerometer) indicating the motion the wearable electronics have undergone. Motion metrics are also described herein as quantifiable. In some embodiments, each motion metric is a single numerical value resulting from a quantified combination (e.g., addition, averaging, etc.) of motion data distributed along one or more axes of the motion sensor. This single numerical value is a concise representation of the user's motion and, as described above, can be used to detect the user's state, sleep stage, activity level, and other types of user states.

[0090] Using a motion metric quantified at a certain interval or frequency (e.g., every 30 seconds), embodiments generate informative statistical characteristics of a user's motion over longer periods (e.g., 10 minutes). In one embodiment, the motion metric measures user movement and behavior over short time intervals (e.g., 30 seconds). More specifically, this metric quantifies and combines the distribution of acceleration data (a type of motion data) along an axis into a single numerical value. Assume a_x, a_y, and a_z represent time series arrays of accelerometer data measured along the three axes, sampled at a certain sampling rate (e.g., 20 Hz) over a certain time interval (e.g., 30 seconds).

[0091] One embodiment of the motion measure uses the maximum and minimum accelerations along the axis to quantify the distribution of acceleration along said axis. Individual values ​​from each axis are summed to form a single measure every 30-second interval (other intervals are also possible). The motion measure generated from a set of motion data produced within a time period can then be expressed as:

[0092] MM = max(a_x) - min(a_x) +

[0093] max(a_y)-min(a_y)+

[0094] max(a_z)-min(a_z)

[0095] The more general form of this measure allows for different weightings along different axes and different exponents for each axis. w_x, w_y, w_z are the weighting factors, and exp_x, exp_y, exp_z are the exponents. The shift measure can be expressed as:

[0096] MM=w_x*(max(a_x)-min(a_x))^exp_x+

[0097] w_y*(max(a_y)-min(a_y))^exp_y+

[0098] w_z*(max(a_z)-min(a_z))^exp_z

[0099] It should be understood that some embodiments may perform several other post-processing operations on the motion metric. For example, in other embodiments, the motion metric generator may perform an operation utilizing saturation (clamping the calculated motion metric to boundary values ​​supported by several bits used to store those calculated values). In some embodiments, the motion metric generator may project the calculated motion metric onto a 4-bit variable (or 8-bit, 16-bit), which may, in some cases, minimize the data volume. In some embodiments, the frequency of occurrence of positively saturated calculated motion metrics may be used to determine a threshold above which the projection is shifted to produce a motion metric (e.g., to account for a "noisier" motion sensor that causes meaningful data to be in a higher order of the calculated motion metric). For example, in an embodiment using 4 bits to represent the motion metric, a 0-shift means that the motion metric is projected from bit position 3: 0 of the calculated motion metric; a 1-shift means that the motion metric is projected from bit position 4: 1 of the calculated motion metric, and so on.

[0100] In one embodiment, the calculated motion measure includes a combination of statistical measures of motion data along one or more axes. The statistical measures of the motion data may be one or more of quantiles (e.g., quantiles of acceleration data), interquartile ranges, measures of deviation (e.g., Gini coefficient), and measures of entropy. Furthermore, in one embodiment, the statistical measures may use information in the frequency domain. Information in the frequency domain can be used to quantify the amount of periodic motion and compare the amount of periodic motion in two or more frequency bands.

[0101] Other methods for quantifying the distribution of motion data (including accelerometer data) are also possible:

[0102] • Use standard deviation instead of the maximum minus the minimum value

[0103] • Use the interquartile range of motion data.

[0104] Other methods for quantifying user movement are also possible, such as:

[0105] • Use the time derivative of motion data along each of the three axes (also known as push, surge, or hover).

[0106] • Measure the deviation between the total acceleration and 1G (gravitational acceleration at sea level on Earth).

[0107] • Integrating accelerometer data (e.g., the first integral is velocity, and the second integral is position).

[0108] Note that although acceleration data measured by an accelerometer is given as one instance of motion data, other sensors (e.g., gyroscopes, gravity sensors, rotation vector sensors, or magnetometers) can be used to collect other types of motion data to produce motion measurements.

[0109] One advantage of using motion metrics is their extremely low cost of operation on embedded devices. For example, in some embodiments, a single value of the motion metric can capture enough meaningful information to perform the determinations described later herein, but requires relatively little power to compute, relatively little storage space to store, and relatively little bandwidth to transmit, etc. Furthermore, a finite number of bits can be used to represent the motion metric (e.g., as described above, a motion metric can be projected onto a 4-bit variable), minimizing the data volume. As discussed above, the motion metric can be passed from the wearable electronics to a computer server, or to an application on the device (e.g., an app on a mobile phone or tablet) for further processing, such as determining the user's state.

[0110] Statistical measures of user mobility

[0111] As previously seen Figure 1 As described in task box 3 shown in the diagram, statistical measurer 122 can perform statistical measurements to generate feature values ​​for a given time. A user's non-instantaneous behavior can be characterized using a distribution (and distribution-derived features) of the moving measure across various time scales or window sizes (e.g., 5, 10, 40-minute intervals). One embodiment of statistical measurer 122 uses a scrolling window centered at the time of interest (the time to which the feature value is assigned). For example, the feature value F is used as a W-minute window centered at time T. N The characteristics of the distribution of the motion measure extending from TW / 2 to T+W / 2 will be described. In other embodiments, the statistical measure 122 may use a rolling window aligned left or right to time T to consider only data occurring before or after time T. The rate at which the feature values ​​are calculated may correspond to the rate at which the motion measure is generated, the window size, or any other suitable rate. For example, the next feature (e.g., F) N+1 This can correspond to time T+W, T+W / X (where X is any value), T+MMI (where MMI represents the interval of the movement measure (e.g., 30 seconds according to some implementations)), or T+X.

[0112] In one embodiment, determining a user's state can utilize multiple sets of statistical features across multiple time windows. Exemplary statistical features are:

[0113] • The rolling fraction of the movement metric is below a threshold (e.g., 0.13g or any other suitable value).

[0114] • The rolling fraction of the movement metric is above a threshold or within a range of the threshold (e.g., above 0.17g or between 0.17g and 0.20g), or any other suitable value.

[0115] • The rolling fraction of the continuous movement metric is above a threshold (e.g., 0.2g), or any other suitable value.

[0116] • Various rolling quantiles, mean, standard deviation and other summary statistics for these distributions.

[0117] Information from optical photoplethysmography (PPG)

[0118] like Figure 1 As described, in optional task box 10, a photoplethysmography (PPG) sensor 114 can optionally be used to generate PPG data to calculate heart rate (HR), HR variability (HRV), and / or respiratory rate (RR). The PPG sensor typically includes a light source and a photodetector. The common light source is a light-emitting diode (LED).

[0119] PPG data can be used to calculate a user's heart rate by measuring the time between peaks or by calculating the dominant frequency in an optical signal. For most users, heart rate decreases shortly after the onset of sleep and continues to decrease during sleep until earlier in the morning. Heart rate increases when the user wakes up or during short disturbances in sleep. Measuring heart rate allows us to better identify sleep cycles. Furthermore, heart rate is a good indicator of separating sleep cycles from resting cycles, which can confuse motion sensors such as accelerometers. Some users do not experience this characteristic decrease in heart rate during sleep. These users can be identified after wearing wearable electronics for several days, and different algorithms can be used for these users.

[0120] The difference between a user's heart rate and their mean resting heart rate can be used as a personalized measure of how low their heart rate has dropped. PPG data can also be used to calculate a user's respiratory rate. Respiratory rate often shows distinct changes during sleep and can be used to identify characteristics of the sleep cycle. Respiration can be calculated using several techniques:

[0121] • Measure the variability in the peak interval between heartbeats

[0122] • Measure the slow variability (0.1-0.5 Hz) in the baseline of the PPG signal.

[0123] • Measure slow periodic signals (0.1-0.5Hz) in acceleration signals.

[0124] In one embodiment, data representing the heart rate, heart rate variability, and / or respiratory rate of a user of the wearable electronics device are calculated by a PPG data analyzer 118, and the resulting data (referred to as “analyzed PPG data”) is additional input to the user activity classifier 124 and / or the multi-class sleep stage classifier 128, in addition to the statistical features obtained via the statistical measurer 122.

[0125] Classified as activity level

[0126] As previously seen Figure 1 As illustrated in task box 4A, user activity classifier 124 can classify statistical features associated with a time period or otherwise label them with activity levels, such as categorizing them as active or inactive. In some embodiments, user activity classifier 124 may comprise a classification model generated by running a supervised machine learning algorithm on a set of sample features with known labels (e.g., a set of features associated with an active state (e.g., the wearer is more likely to wake up) or an inactive state (e.g., the wearer is more likely to fall asleep)).

[0127] In one embodiment, the user activity classifier is a decision tree, but many other classifiers are possible, for example (with the help of examples rather than restrictions):

[0128] Random Forest

[0129] Support Vector Machine

[0130] Neural Networks

[0131] K-Nearest Neighbor Algorithm

[0132] · Bayes

[0133] Hidden Markov Model

[0134] Furthermore, the user activity classifier can use elevation to combine these various machine learning classifier models for more accurate classification. In one embodiment, the user activity classifier can post-process the classification model to remove noise and improve classification using binary opening and binary closing operations. Many other smoothing operations are possible, including median smoothing, weighted smoothing, Kriging, Markov models, etc.

[0135] The accuracy of inferred transition points between user states can be improved using a variety of methods, including edge detection and change point analysis.

[0136] As discussed above, user activity classifiers can use features derived from data other than motion data. These non-motion data-derived features, which can also be incorporated into this classifier, include:

[0137] Heart rate data, such as PPG

[0138] • Changes in skin or ambient temperature over time

[0139] • Skin conductance response

[0140] • Ambient light measurement (e.g., people tend to sleep in dark environments)

[0141] • Sound detection used to detect patterns associated with sleep (e.g., snoring, deep breathing, or relatively little noise).

[0142] In one embodiment, the plurality of user activities includes a third detected state indicating when the device is not worn, and this "non-wearable" state can be used to improve accuracy. This "non-wearable" state can be inferred from statistical characteristics derived from motion sensors, PPG data, and / or from data from sensors such as temperature sensors, ambient light sensors, skin conductance sensors, capacitive sensors, humidity sensors, and sound sensors.

[0143] For example, in the case of a motion sensor only, the "non-wearable" state can be detected when the wearable electronics remain in a specific orientation and / or remain extremely still for too long, as this is not a typical characteristic of human sleep patterns. In the case of a motion sensor + PPG sensor, the "non-wearable" state can be detected from the optical characteristics of the PPG data, possibly combined with a low-motion state (as measured using a motion sensor). Sleep onset and occurrence can also be inferred from the timing and duration of the "non-wearable" state. Sleep estimates can be improved over time for the user in response to user editing of sleep onset and occurrence estimates (by using this feedback as training data or as a priori when generating classification predictions). While the "non-wearable" state can be detected using a motion sensor + PPG sensor, alternative embodiments that do not require a PPG sensor are discussed below.

[0144] Exporting awake and asleep states

[0145] See above. Figure 1 As illustrated in task frame 4B, the time block classifier 126 derives time blocks from an activity level where the user is in one of multiple sleep states within a set of time periods. Sleep states may include awake and asleep states. The time blocks may span one or more time periods assigned to an activity level. Consecutive time blocks may have different sleep states.

[0146] The time block classifier 126 can transition time periods assigned to an active or inactive level to a sleep state based on empirically derived logic. Generally, an active state implies wakefulness and an inactive state implies sleep, but various opposing edge cases during this transition phase should be considered. For example, when there is substantial step activity (>200 steps) immediately before and after a short inactive period (<60 minutes) within 20 minutes, the inactive period can be classified as a "wake-up" time. An example of when this can be helpful is correctly identifying a one-hour meeting during which the user sits quietly as a wake-up time rather than a nap.

[0147] With automatic detection of the user's status, wearable device users no longer need explicit user instructions to plan the transition to a sleep state. This can be inconvenient and error-prone for wearable device users, as users often forget to notify the wearable device of the transition.

[0148] Sleep stage detection

[0149] Return to view Figure 1 The statistical characteristics of activity levels can be modified or also used to classify a user's sleep stages (note the dashed boxes indicating user sleep state detection 190 and sleep stage classification 192). For example, a wearable electronic device may implement user sleep state detection 190 but not sleep stage classification, nor perform this classification in a different manner. As another example, a wearable electronic device may implement sleep stage classification 192, but implement user sleep state detection in a different manner or not at all (e.g., the wearable electronic device may contain an interface for the user to notify the user of a planned transition to a sleep state (e.g., via the user pushing a button on the wearable electronic device or clicking on the wearable electronic device), and respond by enabling sleep stage classification 192 until the user wakes up (the transition back to a waking state may be via notification from the user or via automatic detection (as described herein, or via different technologies)). Therefore, user sleep state detection 190 and sleep stage classification 192 are independent but can be used together. Figure 1 At task box 5, based on the group statistical features (optionally also based on HR / HRV / RR), the multi-level sleep stage classifier 128 can determine the user's sleep stage.

[0150] Human sleep patterns can be described based on discrete sleep stages. At the highest descriptive level, these are divided into REM (rapid eye movement) and non-REM sleep stages, the latter containing various light and deep sleep stages. Movement patterns, heart rate, heart rate variability (HRV), and respiration change according to sleep stages. These patterns can be detected using data from motion sensors (e.g., triaxial accelerometers) along with photoplethysmography sensors that allow measurement of heart rate, HRV, and respiration.

[0151] The data collected from the motion sensor and photoplethysmography sensor includes:

[0152] • Measurement of user mobility (see the discussion of mobility metrics above in this article)

[0153] • Data used to calculate heart rate, heart rate variability, and respiration (e.g., determined using data from a photoplethysmography sensor in an electronic wearable device).

[0154] • Time-domain or frequency-domain measure of heart rate variability.

[0155] Statistical characteristics can be derived using movement measurements, and additionally, raw heart rate data can be used to define several higher-level characteristics. For example:

[0156] • Rolling median or average heart rate on various time scales (e.g., the previous minute, the previous 5 minutes, etc.)

[0157] • Such as the rolling quantiles of heart rate distributions measured on various time scales

[0158] • Whether the heart rate exceeds / falls below a certain threshold during a certain interval.

[0159] Additionally, statistical characteristics can be derived using the time derivative of heart rate. For example:

[0160] • Maximum rolling change of heart rate on various time scales

[0161] • The rolling median, mean, or other quantile of the distribution of heart rate derivatives on various time scales.

[0162] Similar statistical characteristics can be derived using heart rate variability or respiration. For example:

[0163] • The number of times HRV or respiratory events exceed a certain threshold on various time scales.

[0164] • The rolling median, mean, or some other quantile of the distribution of HRV or respiration over various time scales.

[0165] Heart rate, its derivative, heart rate variability, and respiration can be appropriately regularized or standardized on a per-user basis, taking into account natural variations in the user. For example, a user's resting heart rate can be subtracted from their instantaneous heart rate. Demographic information can also be used to re-normalize features to account for variations based on demographic data.

[0166] Subsequently, the derived statistical features can be used to train a multi-level classifier (via supervised machine learning)128 to classify time cycles into specific sleep stages (e.g., REM, various non-REM (NREM) stages, such as light or deep sleep).

[0167] One embodiment uses a random forest classifier. Other classification methods include neural networks, hidden Markov models, support vector machines, k-means clustering, and decision trees. Other change detection methods can also be used to detect transitions between stages, such as change point analysis, t-tests, and Kolmogorov-Smirnov statistics.

[0168] The classification labels generated from previous steps can be smoothed to reduce noise. For example, classifications generated from a random forest classifier can be processed using an empirically calibrated Markov model with specified transition probabilities between states. Other smoothing operations may include binary erosion / amplification, median smoothing, etc.

[0169] It should be noted that Figure 1 The automatic user state detection and user sleep stage detection described herein can be performed in parallel within a single system. User state detection and user sleep stage detection can use the same or different statistical characteristics to make their own determinations.

[0170] Illustrative example of user state detection

[0171] Figure 2A This describes, according to an embodiment of the invention, the collection of motion metrics at different time windows for a time of interest, the determination of statistical characteristics of one of the times of interest, and the classification of the times of interest based on said statistical characteristics. As used herein, "time of interest" may refer to a given point in time, for example, a time period corresponding to a motion metric, statistical characteristic, or state, and therefore, the time of interest may be used interchangeably with any suitable indication of a single point in time or a range of time periods. Motion metrics are collected for each time interval (see...). Figure 1 (Refer to 116 and task box 2). Based on motion metrics and optionally PPG data, within a time window, at reference 212, the embodiment calculates one or more statistical features of the time within the window (see...). Figure 1 (Refer to 122 and task box 3). In one embodiment, the statistical feature may utilize a motion measure from multiple time windows. The motion measure may have discrete values ​​represented by a predefined number of bits. At reference 214, the statistical feature F1-F1 determined for the time of interest is used. N The time of interest is categorized into one of several user activity levels, including active / inactive (see...). Figure 1 (Refer to 124 and Task Box 4A).

[0172] Note that while in one embodiment the calculation of one or more statistical features uses moving measures at consecutive subsets of time intervals falling within a time window containing the time of interest, in alternative embodiments other styles of moving measures (e.g., every other time interval) within said subsets may be used for the calculation.

[0173] Figure 2BAccording to one embodiment of the invention, the awake / sleep phase is derived using the user's active / inactive activity level. As illustrated, the activity level is described as active or inactive over time (at each point of interest). The embodiment derives non-overlapping consecutive time blocks from the active / inactive levels, during which the user is in one of a plurality of states; said states include an awake state and a sleep state, said time blocks span one or more of the points of interest, and said consecutive time blocks have different states among said plurality of states (see [reference]). Figure 1 (Block 126 and task block 4B). Specifically... Figure 2B The following time blocks are derived: Awake 222, Asleep 224, Awake 226, Asleep 228, Awake 230, and Asleep 232; each of these spans multiple moments of interest. With the aid of instances, a given time block can contain moments categorized into different user states (e.g., Asleep 228). Transitions in a user's state are represented by the edges of time blocks (i.e., when one block ends and the next begins—e.g., the user transitions from an awake state to a sleep state at the end of time block 222 / the beginning of time block 224).

[0174] Figure 3A This describes a snapshot of a motion metric obtained from data over different time spans according to an embodiment of the invention. The motion metric is represented by 4 bits, thus having 16 distinct values ​​from 0 to 15 (with an offset of +1). Any value higher or lower than the 15 values ​​is clamped to the highest (15) and lowest (0) value. The time span is between 10:00 PM and 10:00 AM, and the motion metric changes dynamically.

[0175] Figure 3BThis invention describes how, according to one embodiment of the invention, data from different time spans are used to determine statistical characteristics for a time of interest. The characteristic (characteristic 1) is calculated using a rolling window for every 30-second time interval. Characteristic 1 uses a motion measure between the past 20 minutes and the next 20 minutes. For example, the calculation of characteristic 1 at 4:00 AM can be performed as follows: Motion measures generated between 3:40 AM and 4:20 AM are stored in an array. If motion measures are generated at 30-second intervals, then there can be 40 minutes × 2 values / minute + 1 motion measure. The "+1" considers the time of interest corresponding to 4:00 AM. Determinable statistical characteristics include the mean, standard deviation, 75th quantile (or any other suitable quantile), etc. In this example, the statistical characteristic is the number of motion measures greater than a threshold (e.g., 12). For each time of interest, the percentage of motion measures greater than the threshold (e.g., 12 in the above example) within the associated window is recorded as the value of characteristic 1 at the time of interest. In other words, F1 = (the number of movement metrics with values ​​> the threshold (e.g., 12)) / the total number of movement metrics of the window.

[0176] Figure 3C According to one embodiment of the invention, data from different time spans are used to classify user activity levels at a time of interest. As illustrated, based on the statistical characteristics discussed above, user activity is determined to be up to 12:00 AM (reference 241). The user's state transitions to inactivity between 12:00 AM and 4:00 AM (reference 242), thereafter the user remains active for a period of time (reference 243), then returns to an inactive state (reference 244), and subsequently returns to activity at reference 245. The user then undergoes a bi-state transition between inactivity (reference 246) and activity (reference 247). Regarding... Figure 3C The time period covered by the active period 241 can (for example) represent the time of interest (not shown) within the active period 241 as being classified by active state.

[0177] Figure 3D This invention illustrates how data from different time spans is used to derive a user's awake / sleep state according to one embodiment of the invention. For most of the time periods of interest, an active state results in a user being awake, while an inactive state results in a user being asleep. However, Figure 3CThe description derives the sleep state from inactive 242 (representing moments of interest classified as inactive), active 243 (representing moments of interest classified as active), and inactive 242 (representing moments of interest classified as inactive) (refer to 252). That is, for some moments of interest, the user may be classified as active, but the user may still be identified as asleep. If certain criteria are met, the embodiment can classify time spans with "active" states as sleep states. For example, if the user has restless cycles, which intersperse active and inactive moments. The embodiment may employ most rule-based decisions, using the state of individual moments within a time span that includes the previous time span classified as sleep, the subsequent time span having an inactive state, and the current "active" time span having a length within a specific range (10-120 minutes).

[0178] Figure 3E The invention illustrates an embodiment that includes a non-wearable state, using data from different time spans to derive a user state (e.g., active / inactive / non-wearable state). The non-wearable state is indicated at reference 260 (representing the time of interest classified as non-wearable (not shown)), where the embodiment determines that the user was not wearing an electronic wearable device at approximately 9:00 AM.

[0179] Figure 3F The illustration describes how a user's awake / sleep state is derived using data from different time spans, based on an embodiment that considers a non-wearable state. In the illustrated embodiment, the non-wearable state is derived as part of the awake state 275 based on empirically derived logic—at approximately 9 AM, the user is likely to be awake.

[0180] Several embodiments may have rules for assigning sleep states to time blocks based on the wrist-worn state. For example, consider a scenario where a user wakes up and immediately removes their device to shower. Motion metrics may remain low during this time, so this time span would be classified as inactive and ultimately as a sleep period. However, because this time span is classified as non-wearable and the previous state was "sleep" followed by "awake," embodiments can correct for a sleep state by classifying this time block as an awakening period. Conversely, during a user's sleep, the non-wearable criteria are met to some extent. In this case, the period before and after the "non-wearable" time span is considered "sleep." In this case, embodiments can classify the non-wearable period as a sleep period because it is unlikely that the user would actually remove the device during the sleep cycle.

[0181] Illustrative example of user sleep stage detection

[0182] Figure 4AAccording to one embodiment of the present invention, statistical characteristics are determined at different time windows for a time of interest, and statistical characteristics of one of the times of interest are determined and the times of interest are classified based on said statistical characteristics. Figure 4A Similar to Figure 3A And the same or similar reference indicates elements or components with the same or similar functionality. The difference is that the statistical characteristics may differ from the statistical characteristics used to determine the user's state. Statistical characteristics F'1 to F' are determined for the time of interest. M Used to determine the sleep stage at the time of interest.

[0183] Figure 4B This describes the results of determining the sleep stage at different time windows according to one embodiment of the present invention. In this embodiment, there are three NREM stages. Other embodiments may have different NREM stages.

[0184] Flowchart for motion measurement generation, user state detection, and user sleep stage detection

[0185] Figure 5 A flowchart illustrating the generation of motion metrics and optional user state detection and / or user sleep stage detection according to an embodiment of the present invention is provided. Method 500 may be implemented in a wearable electronic device, an electronic device coupled to said wearable electronic device (e.g., a server (containing hardware and software) / tablet computer / smartphone containing an application (referred to herein as an "app")), or distributed between said two (e.g., references 502 and 504 may be implemented in the wearable electronic device, while references 506 and 508 may be implemented in an electronic device coupled to the wearable electronic device).

[0186] At reference 502, motion data generated by a motion sensor is received. This motion data may, in some cases, be accelerometer data representing motion along three axes received from an accelerometer (e.g., a triaxial accelerometer). In alternative embodiments, the motion sensor is a gyroscope, gravity sensor, rotation vector sensor, position detection device (e.g., a GPS device or a device capable of measuring movement using a cellular phone or WiFi triangulation), or magnetometer. While in one embodiment the motion sensor is in a wearable electronics device that is worn on the user's body (e.g., arm, ankle, or chest) or embedded in the user's clothing, alternative embodiments may have motion sensor data generated by another external electronic device and received by the wearable electronics device. The motion data may comprise multiple samples generated during different time intervals, such as a first time interval and a second time interval.

[0187] At reference 504, the wearable electronics generate a measure of motion for each time interval (e.g., every 30 seconds) based on a quantized combination of motion data generated during corresponding time intervals. For example, in the case of acceleration data received from a triaxial accelerometer, the wearable electronics can quantize the acceleration data along each of the three axes during a first time interval and subsequently along each of the three axes during a second time interval. Thus, multiple samples of acceleration data received from two different time intervals are each quantized into a single numerical value. Different techniques for quantizing multiple samples of motion data have been described above.

[0188] Furthermore, it should be understood that the movement measure of the time interval can be a single numerical value. Moreover, this single numerical value can be represented by a predefined number of bits (e.g., four bits). In one embodiment, the number of bits is chosen such that it is small enough to facilitate transmission between different modules within the wearable electronics or between the wearable electronics and electronics coupled to the wearable electronics, and large enough to represent sufficiently meaningful data for the subsequent operations described herein.

[0189] Optionally at reference 506, sleep states can be assigned to time blocks based on motion metrics. These states may include awake and asleep states. Successive time blocks may refer to different sleep states, such as several asleep states for a first time block and awake states for a second time block. As discussed herein, sleep states can be determined by generating features for a given time point based on a motion-metric-based window. These features can then be used to assign activity levels to the given time points. Activity levels are then used to assign sleep states to the time blocks.

[0190] Optionally, at reference 508, for a time of interest, one of a plurality of sleep stages of the user is determined. Each of the times of interest corresponds to one of a time interval, and the determination of each of the times of interest is based on a measure of movement from a subset of time intervals falling within a time window containing the time of interest. The plurality of sleep stages includes a rapid eye movement (REM) stage and a plurality of non-REM stages.

[0191] Motion measurements and / or determined time blocks with the stated state and / or sleep phase may be presented to the user (e.g., on the display of the device or another electronic device (e.g., a tablet / smartphone / computer that receives data from the wearable electronics, generates the data, or receives data from another electronic device (e.g., a server)) or stored in the wearable electronics for a period of time sufficient to present or transmit this data to a secondary device.

[0192] It should be noted that the quantification of the distribution of motion data can be achieved through a combination of methods as discussed above in this article.

[0193] Flowchart for detecting a user's sleep cycles and sleep stages

[0194] Figure 6 This is a flowchart illustrating a method 600 for automatically detecting a user's sleep cycle according to an embodiment of the present invention. Method 600 is... Figure 5 The exemplary implementation of reference 506. Method 600 may be implemented in a wearable electronic device, an electronic device coupled to said wearable electronic device (e.g., a server (containing hardware and software) / tablet computer / smartphone containing an application (referred to herein as an "app")), or distributed between said two (e.g., references 602 and 604 may be implemented in the wearable electronic device, while references 606 and 608 may be implemented in an electronic device coupled to the wearable electronic device).

[0195] Reference 602 describes receiving motion measurements for said time intervals based on a quantized combination of the distribution of motion data along each of the three axes for each of a plurality of time intervals. Motion data representing motion along the three axes is generated by motion sensors (which may be in the wearable electronics or provided to the wearable electronics by another electronic device). In one embodiment, the three axes may be orthogonal axes of the wearable electronics. Motion measurements may be generated by one or more multiple motion sensors, for example, as described above herein. Figure 1 and Figure 5 Various ways of elaboration arise.

[0196] Referring to 604, a set of one or more statistical features is determined for each of a plurality of moments of interest, wherein at least one characterizes the distribution of a motion metric determined for a subset of time intervals falling within a time window containing the moments of interest. The moments of interest may refer to a particular time period of interest and may be used interchangeably with the phrase “a time period.” Each of the moments of interest corresponds to one of the time intervals, and the set of statistical features characterizes the user’s motion. In some embodiments, the time window is a scrolling window centered, left-aligned, or right-aligned relative to the moments of interest. By way of example, the time window may be a variety of time scales, such as 5, 10, or 40-minute intervals, while the time intervals of the motion metric are at smaller time scales, such as 10, 30, or 60 seconds. In one embodiment, the subset of selected time intervals is consecutive time intervals. In alternative embodiments, subsets of time intervals are selected in different patterns, such as every other time interval or every two time intervals. In one embodiment, multiple time windows at different time scales may be used to determine one or more of the statistical features.

[0197] Reference 606 describes classifying a time point into one of a plurality of user states (e.g., activity levels) based on the set of statistical characteristics determined for each of the plurality of time points of interest, wherein the user's state includes active and inactive. In one embodiment, the classification is further based on PPG data generated by a PPG sensor (e.g., a wearable electronic device or another electronic device that transmits the PPG data to the wearable electronic device). For example, at least one or more of the user's heart rate data, heart rate variability data, and / or respiratory data (i.e., analyzed PPG data) are calculated using PPG data from a time window containing the time points of interest. Reference 606 may be related to the above description regarding... Figure 1 Implement in the various ways described in reference 124 and task box 4A.

[0198] Reference 608 describes deriving non-overlapping consecutive time blocks from the user's state at the time of interest, during which the user is in one of multiple states. The states include waking and sleeping states, and each of the time blocks spans one or more of the times of interest, with consecutive blocks having different states. Reference 608 can be understood from the above description regarding... Figure 1 Implement in the various ways described in reference 126 and task box 4B.

[0199] Figure 7 This is a flowchart illustrating the automatic detection of a user's sleep stage according to an embodiment of the present invention. Method 700 is... Figure 5 The exemplary implementation of reference 508. Method 700 may be implemented in a wearable electronic device, an electronic device coupled to said wearable electronic device (e.g., a server (containing hardware and software) / tablet computer / smartphone containing an application (called an app)), or distributed between said two (e.g., references 702 and 704 may be implemented in the wearable electronic device, while reference 706 may be implemented in an electronic device coupled to the wearable electronic device).

[0200] Referring to 702, a quantized combination of motion data distributions along each of the three axes for each of multiple time intervals is received for said time interval. In implementation... Figure 6 and Figure 7 In both embodiments, references 602 and 702 may be the same.

[0201] Reference 704 describes defining a set of one or more statistical features for each of a plurality of moments of interest, wherein at least one of them characterizes the distribution of a motion metric determined for a subset of time intervals falling within a time window containing the moments of interest. Each moment of interest corresponds to one of the time intervals, and the set of statistical features characterizes the user's motion. Reference 704 may be related to the above description regarding... Figure 1 Implement in the various ways described in reference 122 and task box 3.

[0202] Reference 706 describes classifying the moments of interest into one of a plurality of sleep stages of the user based on the set of statistical characteristics determined for each of the plurality of moments of interest. The sleep stages include a rapid eye movement (REM) stage and a plurality of non-REM stages. In one embodiment, the classification is further based on photoplethysmography (PPG) data generated by a photoplethysmography sensor in a wearable electronic device. For example, at least one of the user's heart rate data, heart rate variability data, and respiratory data is calculated using PPG data from a time window containing the moments of interest. Reference 706 may refer to the above regarding... Figure 1 Implement in the various ways described in reference 128 and task box 5.

[0203] Note that the sleep stage detection in method 700 can be performed after it is determined that the user is asleep, and the determination of being asleep can be achieved in different ways (e.g., via method 600, via the user interacting with an interface on the wearable electronic device (e.g., pushing a button or clicking the shell)).

[0204] Automatic power consumption change of photoplethysmography / motion sensor

[0205] In embodiments that can automatically detect a user's state and / or sleep stage, this information can be used to adjust the power consumption of various sensors, such as photoplethysmography (PPG) sensors and motion sensors. For example, accurate measurement of heart rate variability requires greater temporal accuracy in the signal compared to the temporal accuracy necessary to measure heart rate; therefore, the PPG sensor's light source can be measured at a higher sampling rate and / or higher power for a fraction of a sleep cycle (and / or wake-up cycle) to achieve a better estimate of heart rate variability. Similarly, the accuracy of measurements of motion along an axis can be improved with higher sampling rates, sensitivity, and / or power levels, which can help determine, for example, a user's sleep stage.

[0206] Conversely, once the user is detected to be asleep, the embodiments can reduce the power consumption of the photoplethysmography sensor and / or motion sensor.

[0207] Figure 8AA flowchart illustrating the automatic reduction of power consumption of at least one of a photoplethysmography (PPG) sensor and a motion sensor according to an embodiment of the present invention is provided. The method can be implemented by a wearable electronic device comprising a set of sensors, including a motion sensor (e.g., an accelerometer) and a PPG sensor (which contains a light source and a photodetector (e.g., an LED)). At reference 802, in response to a user's state transition, such as being tracked by the wearable electronic device, to a sleep state, the wearable electronic device reduces the power consumption of at least one of the PPG sensor and the motion sensor. Optionally then, at reference 804, following this reduction, the wearable electronic device periodically and temporarily increases the power consumption of the at least one of the PPG sensor and the motion sensor to generate additional data for sleep stage detection. At reference 806, in response to a user's state transition, such as being tracked by the wearable electronic device, to leave a sleep state, the wearable electronic device reverses the reduction in power consumption of the at least one of the PPG sensor and the motion sensor.

[0208] In one embodiment, the reduction in power consumption includes one of the following: a reduction in the sampling rate of the photoplethysmography (PPG) sensor, a reduction in the sensitivity of the PPG sensor, a reduction in the power level of the PPG sensor's light source, entry into a low-precision state of the motion sensor, a reduction in the sensitivity of the motion sensor, and a reduction in the sampling rate of the motion sensor. In one embodiment, the motion sensor is an accelerometer.

[0209] In one embodiment, the temporary increase in power consumption includes at least one of the following: an increase in the sampling rate of the photoplethysmography sensor, an increase in the sensitivity of the photoplethysmography sensor, an increase in the power level of the light source, entering a high-precision state of the motion sensor, an increase in the sensitivity of the motion sensor, and an increase in the sampling rate of the motion sensor.

[0210] Figure 8B A flowchart illustrating the power consumption of at least one of a photoplethysmography sensor and a motion sensor according to an embodiment of the present invention.

[0211] At reference 812, in response to a user's state transitioning to a sleeping state, as tracked by the wearable electronics, the wearable electronics increase the power consumption of at least one of the photoplethysmography (PPG) sensor and the motion sensor. As previously described, the increased power consumption provides additional PPG data or motion data for sleep stage detection.

[0212] At reference 814, in response to a change in state of a user, such as when the wearable electronic device tracks a change from a sleeping state, the wearable electronic device reverses the increase in power consumption of at least one of the photoplethysmography sensor and the motion sensor.

[0213] Although in one embodiment targeted Figure 8A The method described in 8B, where user state tracking is performed by the wearable electronic device, is actually performed by the wearable electronic device itself. However, in alternative embodiments, user state tracking is assisted by one or more other electronic devices (e.g., a server (comprising hardware and software) / tablet / smartphone with an application (called an app) that transmits the user's state back to the wearable electronic device in real-time or near real-time; and optionally, sensors external to the wearable electronic device (e.g., the user's chest, mattress, or bedside table) that provide data to the server / tablet / smartphone to aid in detecting transitions into and out of sleep). Both methods detect the user's state and are considered within the meaning of "as tracked by a wearable electronic device."

[0214] While in one embodiment power consumption changes in response to the detection of a user entering and leaving a sleep state, alternative embodiments instead or additionally ensure that power consumption changes in response to the detection of a transition between one or more sleep stages.

[0215] Changes in the power consumption of a sensor can cause visible changes in wearable electronics. For example, a decrease in the power consumption of a photoplethysmography (PPG) sensor can cause the light source to emit less light (reducing its illumination) when the power level of the light source decreases. Similarly, an increase in the power consumption of a PPG sensor can cause the light source to emit more light (increasing its illumination) when the power level of the light source increases. Therefore, wearable electronics can change their illumination immediately upon transitioning between entering and leaving a sleep state.

[0216] An exemplary device implementing automatic detection and automatic power consumption changes relative to motion metrics and user sleep cycles / stages.

[0217] As previously described, while in some embodiments the operations are implemented in wearable electronic devices, alternative embodiments distribute the differences in the operations across different electronic devices. Figure 9 (Illustrate an instance of this type of distribution). Figure 9 This is a block diagram illustrating a wearable electronic device and an electronic device that implements the operations disclosed in an embodiment of the present invention. The wearable electronic device (WED) 902 includes a processor 942, which may be a set of processors. The WED 902 includes a motion sensor 112, which may be a multi-axis accelerometer, gyroscope, gravity sensor, rotation vector sensor, or magnetometer as discussed above herein. The WED 902 may also include other sensors 914, which may include… Figure 1The wearable electronics 902 includes a photoplethysmography sensor 114, a temperature sensor 921, an ambient light sensor 922, a skin conductance sensor 923, a capacitive sensor 924, a humidity sensor 925, and a sound sensor 926. The wearable electronics 902 also includes a non-transitory machine-readable storage medium 918 containing instructions for implementing the motion measurement generator 116 described above. When executed by the processor 942, the motion measurement generator causes the wearable electronics 902 to generate motion measurements. In one embodiment, the non-transitory machine-readable storage medium 918 includes a power consumption regulator 917 that performs power consumption regulation on at least one of the motion sensor 112 and the PPG sensor 114 described above herein.

[0218] In one embodiment, the other sensors 914 are not located within the wearable electronics 902. These sensors may be distributed around the user. For example, these sensors may be placed on the user's chest, mattress, or bedside table, while the wearable electronics are worn by the user.

[0219] Figure 9 It also includes an electronic device 900 (e.g., a server (containing hardware and software) / tablet computer / smartphone running an application (referred to as an app)). The electronic device 900 can perform functionality related to a statistical analyzer 122, a user activity classifier 124, a time block classifier 126, and / or a multi-level sleep stage classifier 128, some or all of which are contained in a sleep tracking module (STM) 950 stored in a non-transitory machine-readable storage medium 948. When executed by the processor 952, the STM 950 causes the electronic device 900 to perform the corresponding operations discussed above herein. The electronic device 900 may contain virtual machines (VMs) 962A to 962R, each of which can execute a software instance of the STM 950. A super supervisor 954 can present a virtual operating platform for the virtual machines 962A to 962R.

[0220] Extended and enabled applications

[0221] A variety of applications and extensions related to sleep are possible using multiple sensors.

[0222] Data from wearable electronics when placed on or near the bed.

[0223] Even when not physically worn by the user, wearable electronics can be used to monitor sleep and the onset / occurrence of sleep stages. For example, an electronic device placed on, on, or within a bedside table can measure a user's sleep. This device will use an accelerometer, microphone, ambient light sensor, or camera to measure the user's movement, heart rate, or breathing. The camera may be functional in illuminating the room and capturing video in low-light or infrared conditions. These features can then be used to infer details of the user's sleep and sleep stages, as described herein.

[0224] Similarly, wall-mounted devices such as alarm clocks can be equipped with cameras and microphones to measure a user's heart rate, respiratory rate, and movement to infer details about sleep. Because the user is asleep, they will be extremely still, providing a good opportunity to estimate heart rate from video. This camera may be functional in lighting the room and capturing video in low-light environments or infrared conditions. Furthermore, this camera can simultaneously capture the heart rate, respiratory rate, and movement of multiple users. Therefore, motion measurements and PPG data can be generated by sensors and data collection devices external to the wearable electronics, and this motion measurement and PPG data can be provided to the electronics, either directly or via the wearable electronics or via more direct transmission (e.g., with its own WiFi, Bluetooth, or cellular connectivity), to infer details of the user's sleep and sleep stages as described herein.

[0225] Because the user is at rest during sleep, this provides the system with the opportunity to measure higher-level parameters about the user, such as minor changes in respiratory rate or (cardiac) heart rate intervals. This information can be used to help detect or diagnose conditions such as sleep apnea and atrial fibrillation.

[0226] Sleep apnea detection

[0227] Sleep apnea can be described as an interruption of normal breathing during sleep. Given automatic sleep monitoring, sleep apnea can be detected in multiple ways using various combinations of sensors. For example:

[0228] • Decreased blood oxygenation using pulsed oxygenation methods (e.g., as part of a PPG system that utilizes PPG data);

[0229] • An interruption of a normal audible breathing pattern as detected by an audio sensor

[0230] • Changes in respiratory rate as measured using a PPG system.

[0231] • Changes in breathing using strain gauges (e.g., worn around the chest).

[0232] • Changes in respiration using an accelerometer (which measures the periodic acceleration of respiration).

[0233] • Changes in breathing observed using cameras that directly or indirectly monitor breathing

[0234] • Changes in respiration were detected using a CO2 sensor that detects changes in the amount of carbon dioxide exhaled.

[0235] Multi-user sleep tracking

[0236] It can track sleep patterns (e.g., sleep onset / occurrence, wakefulness, sleep stages) for multiple users simultaneously. For example:

[0237] • The image recording device located on the bedside table can directly detect sleep by observing the movement of one or more users, whether they are breathing in a way that is consistent with sleep, and by measuring their heart rate and heart rate variability by detecting changes in the color of their skin with each heartbeat.

[0238] • The image recording device can also directly detect the occurrence or absence of rapid eye movements in one or more sleep users, thus directly detecting the difference between REM and non-REM sleep stages.

[0239] A smart mattress with one or more accelerometers can detect and separate movements from multiple sleeping users.

[0240] • The audio sensor can detect and interpret breathing or snoring patterns for multiple users simultaneously.

[0241] Detecting the cause of sleep interruption

[0242] Some sleep disturbances can occur due to physiological (e.g., sleep apnea) and / or environmental influences. These are detectable and associated with sleep disturbances, which can be used by the user to identify things or events that degrade his / her sleep. For example:

[0243] Temperature sensors (worn on the body, manufactured as part of the bed, manufactured as part of a device located on a bedside table, etc.) can detect changes in temperature that can disrupt sleep.

[0244] • The audio sensor can detect sounds that can interrupt or degrade a user's sleep.

[0245] • Ambient light sensors can detect bright, persistent, intermittent light that can disrupt sleep.

[0246] • Humidity sensors can detect changes in humidity that can lead to discomfort or degraded / disrupted sleep.

[0247] Accelerometers can be used to detect movements that can disrupt sleep (e.g., a large truck driven by an earthquake).

[0248] Automatic detection based on motion sensors to determine when wearable electronic devices are not being worn.

[0249] Some embodiments discussed herein relate to wearable electronics capable of detecting when the wearable electronics are not being worn. For example, depending on the arrangement between the housing and a possible wristband, the wearable electronics can detect when the wearable electronics are positioned in one of several orientations typically found when the wearable electronics are not being worn. In some cases, these orientations can be specified by a non-wear profile. As used herein, a “non-wear profile” can be data or logic specifying a pattern of motion data indicating when the wearable electronics are not being worn by a user. In a given embodiment, the pattern of motion data (e.g., accelerometer data) can reflect gravity detected by an inertial sensor (e.g., an accelerometer) along one or more axes.

[0250] While embodiments are described with reference to a triaxial accelerometer oriented in a particular manner in wearable electronics, alternative embodiments may have different orientations, which may require predictable changes to the techniques described herein, such as transforming motion data from one coordinate system to another. For the sake of simplicity in the discussion of the exemplary embodiments, the negative direction of an axis is referred to as the axis's reverse direction.

[0251] By way of example rather than limitation, a specific orientation of the motion axis detected by the accelerometer is now described relative to the display. Consider a wearable electronics device worn on a user's forearm in the same position as the watch's display would be worn (relative to the watch face): the X-axis runs along a line formed between the 12 and 6 o'clock positions (positive direction from 12 to 6) and can also be called the top-bottom axis; the Y-axis runs along a line formed between the 9 and 3 o'clock positions (that is, from the user's elbow to the wrist if worn on the left hand) (positive direction in some cases from 9 to 3) and can also be called the left-right axis; the Z-axis runs along a line perpendicular to the watch face (positive direction from the front of the watch face) and can also be called the back-front axis. Therefore, in this example, the XY axes form a plane containing the display / watch face, and the XZ axis forms a plane perpendicular to the user's forearm.

[0252] In one embodiment, the wearable electronics to be worn on a user's forearm has a housing containing electronic components associated with the wearable electronics, one or more buttons for user interaction, and one or more displays accessible / visible via the housing. According to one embodiment of the invention, the wearable electronics may also include a wristband to secure the wearable electronics to the user's forearm. As used herein, the term "wristband" may refer to a band designed to completely or partially encircle a person's forearm near the wrist joint. The band may be continuous, for example without any interruption (it may be stretchable to fit on a person's hand or have an extension similar to a fashion watch strap); or it may be discontinuous, for example having a clasp or other connection that allows the band to close like a watch strap; or it may simply open, for example having a C-shape that snaps onto the wearer's wrist.

[0253] Some instance orientations will now be described in more detail.

[0254] Figure 10A -E illustrates the orientation in which different exemplary wearable electronic devices can be placed when not worn, according to the embodiments, which can be represented by different non-wearable profiles. Figure 10A -B describes a wearable electronic device with a C-shape and an integrated housing and wristband according to an embodiment of the present invention. Figure 10C -D describes a wearable electronic device according to an embodiment of the present invention, which has a wristband similar to a fashion watch strap. Figure 10E This illustration depicts wearable electronic devices worn by a user as they participate in various activities, according to an embodiment. Note that each wearable electronic device depicted in this set of figures may include an accelerometer.

[0255] Figure 10A This describes a wearable electronic device according to an embodiment of the invention, in which its sides rest on a flat surface when not worn. When placed in this orientation, the left-right axis (Y-axis) is generally parallel to the extension of gravity, and the acceleration along the left-right axis attributable to gravity will always satisfy a condition relative to a threshold acceleration expected along said axis when the wearable electronic device is in this orientation. Although Figure 10A This describes a wearable electronic device on one of its sides, but the wearable electronic device can also typically be placed on its opposite side (e.g., side 1004). Regardless of where the wearable electronic device is placed... Figure 10A In the example shown, whether on side 1004 or side 1004, gravity extends along the left-right axis but in the opposite direction. Therefore, when measured from left to right in the positive direction of the accelerometer, the acceleration data along the left-right axis is displayed on wearable electronic devices such as... Figure 10AAs shown, 1g can be measured when the device is oriented, but -1g can be measured when the wearable electronics are oriented so that they are on side 1004. Because the wearable electronics can be placed on either side, the wearable electronics can compare the detected acceleration data with either of two different thresholds (one for each side, for example, 1g for...). Figure 10A The static side shown in the figure, and -1g is used for the side 1004) comparison, or the absolute value of the acceleration data from the left-right axis is taken, which is then compared with the gravity threshold (e.g., 1g).

[0256] Figure 10B The description describes a wearable electronic device positioned face-up on a flat surface when not being worn, according to one embodiment. When placed in this orientation, the rear-front axis (Z-axis) is generally parallel to the gravitational extension, and the acceleration along the rear-front axis attributable to gravity satisfies a condition relative to a threshold acceleration expected along the axis when the wearable electronic device is in this orientation.

[0257] Note that in one embodiment, due to the shape and flexibility of the C-shaped wristband, it is extremely unlikely that the wearable electronics would be placed with its back side off when not being worn. Therefore, Figure 10A -B indicates the orientation in which wearable electronics are typically placed when not being worn.

[0258] Figure 10C According to one embodiment, when not being worn, the wearable electronics are placed on a flat surface such that the front of the wearable electronics is perpendicular to the flat surface. When placed in this orientation, the top-bottom axis (X-axis) is generally parallel to the gravitational extension, and the acceleration along the top-bottom axis attributable to gravity will always satisfy a condition relative to a threshold acceleration expected along the axis when the wearable electronics are in this orientation.

[0259] Figure 10D According to one embodiment, when not being worn, the wearable electronics are positioned such that the back is on a flat surface. When positioned in this orientation, the rear-front axis (Z-axis) is generally parallel to the gravitational extension, and the acceleration along the rear-front axis attributable to gravity will always satisfy a condition relative to a threshold acceleration expected along the axis when the wearable electronics are in this orientation. Figure 10D The physical characteristics of wearable electronic devices allow them to be typically placed in Figure 10B Within the same orientation described in the text. Compare. Figure 10D and Figure 10B The rear-to-front axle extends in opposite directions relative to gravity. Therefore, the positive direction extends from rear to front and is used for... Figure 10D In the embodiment where the condition is whether the acceleration data is greater than the gravity threshold, logically used for... Figure 10BThe condition will be whether the acceleration data is less than the gravity threshold. Of course, in other embodiments, this logic can be implemented in different ways, for example, by taking the absolute value of the acceleration data and comparing the absolute value with the gravity threshold.

[0260] In one embodiment, Figure 10A The threshold acceleration in -B is based on the force required to counteract gravity on a stationary object and an acceptable degree of tilt; this tilt can be one or more of the orientations in which the wearable electronics are typically placed when not worn, and on an object on which the wearable electronics are placed. For example, in some embodiments, the threshold acceleration is equal to 1g multiplied by a cosine (X°), where X° is selected from the range of 0-40°, and more precisely, the range of 25-35°, and in one particular embodiment for at least Figure 10A The orientation is 30°. While embodiments may use the same degree for all orientations in which the wearable electronics are typically placed when not worn, different embodiments may select different tilt angles for different orientations.

[0261] Figure 10A -D indicates several common orientations in which the wearable electronics can be positioned when not worn, and in one embodiment, these orientations depend on the characteristics of the wearable electronics, including one or more of the following: shape, flexibility, and range of motion of the mechanism (e.g., a wristband) for wearing the wearable electronics on the user's forearm. Furthermore, in one embodiment, these orientations are a result of the physical characteristics of the wearable electronics forming direct physical contact with the object on which it is placed (e.g., a flat surface). For example, although... Figure 10A -D indicates the orientation of a wearable electronic device on a flat surface, but some wearable electronic devices may be designed to be placed differently when not worn, such as in a watch case with an inclined surface. Therefore, the common orientation of a wearable electronic device when not worn is device-specific, and different embodiments will take this common orientation into account for a given device when selecting one or more axes along which gravity will cause the acceleration along those axes to always satisfy a condition relative to a threshold acceleration.

[0262] Compared to the common orientation that wearable electronic devices can be placed when not being worn, Figure 10EThis illustration describes the orientation of a wearable electronic device when a user engages in various activities according to an embodiment of the invention. As illustrated, user 800A is engaged in vigorous physical activity (e.g., tennis), user 800B is walking, user 800C is running, user 800D is practicing yoga, user 800E is sleeping, user 800F is engaged in recreational physical activity (e.g., golf), user 800G is cycling, user 800H is walking a pet, and user 800I (a pet) is walking with its owner. All users are wearing wearable electronic device 100, and all users except the pet wear the wearable electronic device 100 on their forearms. The pet wears the wearable electronic device on its leg. Therefore, the wearable electronic device is designed to be worn in a specific manner, and said specific manner is unlikely to include a sustained threshold amount of time in which the wearable electronic device remains in the one or more orientations that it would normally be in when not worn.

[0263] Return to view Figure 10E When the user is active (e.g., users 800A-800C and 800F-800I), the acceleration measured by the accelerometer changes dynamically over time, making it easier to distinguish from a non-wearable state. However, when the user is nearly stationary (e.g., users 800D and 800E), the acceleration measured by the accelerometer changes very little and / or infrequently, making it more difficult to distinguish from when the wearable electronics are not being worn.

[0264] However, even when the user is relatively stationary, it may be uncommon for the user to maintain the wearable electronics in the same / similar orientation as the wearable electronics would be in if it were not worn. For example, suppose wearable electronics 100 has Figure 10A Similar characteristics to the wearable electronic device described in section -B. When user 800D is asleep, wearable electronic device 100 is beside his body. Therefore, for example, when asleep, the wearable electronic device will be in... Figure 10A The sustained long-duration period described herein (on its side) is extremely unlikely. Therefore, the wearable electronics 100 can be determined to be in a non-wearable state when the acceleration measured along the left-right axis meets a condition relative to a threshold acceleration expected along said axis. For example, said condition could be that the acceleration along the left-right axis exceeds 0.7g or is less than -0.7g for a determined time period (e.g., exactly 5 minutes).

[0265] The conditions used to determine the non-wearable state can be selected based on the probability that the wearable electronics, when worn, will be in the same orientation as it would normally be when not worn. For example, the wearable electronics maintain a certain orientation when not worn. Figure 10B This is not common in orientation, but it is related to Figure 10A Orientation (e.g., the user 800D can sleep while the wearable electronics face downwards for a period of time) is more common in this context. Therefore, the condition used to determine the non-wearable state could be that acceleration along the rear-front axis is below -0.7g for a full 30 minutes.

[0266] Flowchart for automatically detecting when wearable electronic devices are not being worn based on accelerometers.

[0267] Figure 11 This is a flowchart illustrating an embodiment of automatically detecting when a wearable electronic device is not being worn based on motion data. Figure 11 Method 1100 shown can be implemented in a wearable electronic device or distributed between a wearable electronic device and another electronic device. The electronic device can be a server, tablet computer, smartphone (running an application (referred to as an app)), desktop computer, laptop computer, set-top box, or any other computer device or computer system coupled to the wearable electronic device. In one example, references 1102 and 1108 can be executed on the WED, reference 1104 can be executed on the WED and / or another electronic device, and reference 1106 can be executed on another electronic device.

[0268] The operation of method 1100 is now described. At reference 1102, motion data is obtained via a motion sensor of a wearable electronic device. As discussed above, the motion data may be accelerometer data (e.g., acceleration data along a single axis or along multiple axes), and the motion sensor may be an accelerometer (or multiple accelerometers), a gyroscope, a gravity sensor, a rotation vector sensor, a position detection device (e.g., a GPS device, or a device capable of measuring movement using a cellular phone or WiFi triangulation), or a magnetometer.

[0269] At reference 1104, wearable electronics can automatically determine the time periods during which the wearable electronics have not been worn based on a comparison of motion data with a non-wear profile, the non-wear profile specifying a pattern of motion data indicating when the wearable electronics have not been worn by the user. In some cases, the non-wear profile may include a threshold representing gravity along an axis when the wearable electronics are stationary in a given orientation.

[0270] In one embodiment, the time span must be a minimum time length and is variable in length; alternative embodiments implement a fixed-length time span. In one embodiment, the wearable electronics are designed to be worn such that the display of the wearable electronics is in the same position as the display of a wristwatch would be, the axis being parallel to the left-right axis of the display, and the orientation in which the wearable electronics are typically placed when not worn such that the left-right axis is substantially parallel to the Earth's gravity. A single accelerometer for making automatic determinations is along one axis (also referred to herein as a single axis), but the accelerometer may be able to measure acceleration along multiple axes, and alternative embodiments may operate in different ways. For example, in certain of these alternative embodiments, the accelerometers for making different determinations are along different axes (accelerometers along the axis are considered independently, and the differences in the automatically determined time span will be based on the accelerometer along only one of the axes). As another example, in certain alternative embodiments, the acceleration measures used to make at least one of the automatic determinations are along two or more of the axes (accelerations along two or more of the axes are considered collectively, and the automatically determined time span is based on acceleration measures along only two or more of the axes (also referred to herein as the collective axis). Different embodiments may implement different combinations of one or more single axes and collective axes; for example: 1) a single axis will detect the time span in a first orientation, and the collective axis is used for a second orientation; and 2) multiple single axes are used for the first and second orientations respectively, and the collective axis is used for a third orientation.

[0271] Optionally, at reference 1106, the wearable electronics can assign user states to time blocks based on the activity level assigned to the time span. Instances of user states include an awake state and a asleep state. In some cases, successive time blocks have different states. In one embodiment, assigning a time span to a non-wearable active state may cause the wearable electronics to assign an awake state to a portion of a time block containing said time span. However, other embodiments may operate in a different manner. For example, assigning a time span to a non-wearable active state may cause the wearable electronics to assign a asleep state to a portion of a time block containing said time span.

[0272] While in one embodiment reference 1106 may be performed as previously described herein (and thus, automatic detection of when wearable electronics are not worn based on accelerometers may be used in conjunction with the sleep tracker described earlier herein), reference 1106 may be performed using other sleep / wake state detection techniques.

[0273] Optionally, at reference 1108, adjustments are made to one or more of the following based on a time span: WED power consumption, sensor data storage, data transmission / reception, and firmware upgrade scheduling. For example, in one embodiment, WED power consumption is reduced during at least a portion of the time span during which the wearable electronics are not worn (e.g., by reducing its sensitivity or completely powering it off, thus reducing the power consumed by one or more sensors). As another example, in one embodiment, the WED uses a portion of the time span to upload and / or download data from another electronic device.

[0274] Figure 12 Description for implementation from Figure 11 An exemplary alternative embodiment of frame 1104. (See also...) Figure 12 Exemplary alternative embodiments are described, but it should be understood that other alternative embodiments are within the scope of the invention.

[0275] Reference 1204 describes recording a time period as either a worn or unworn state based on acceleration data measured along the accelerometer axis exceeding a threshold acceleration for each consecutive time period. In one embodiment, reference 1204 is performed by a wearable electronic device. While consecutive time periods are non-overlapping in one embodiment, some overlap may exist in alternative embodiments. While the time period is of fixed length in one embodiment of the invention, alternative embodiments may support variable lengths. While the time period is 30 seconds in one embodiment of the invention, alternative embodiments may select a time period in the range of, for example, 5-120 seconds. During a given period of time, the one or more samples from the accelerometer along the axis are collectively represented by recording a single state (worn or unworn) for the time period. This reduces the volume of data used for processing, storage, and / or transmission. Thus, the length of the time period is chosen to sufficiently reduce the data volume while retaining enough meaningful data for subsequent operations described herein. See also Figure 14A The exemplary embodiments are further discussed.

[0276] Furthermore, while in one embodiment time and state are recorded, in an alternative embodiment the time and recorded state are compressed to form data representing the state (e.g., in the case where the time period is of fixed length, the data may contain a start time followed by a stream of 1s and 0s each representing the state during one of the time periods; additional compression (e.g., run-length encoding) may also be performed).

[0277] Regarding the exemplary use of one or more single axes and collective axes discussed above, in one embodiment, a separate state is recorded for each of the one or more single axes and collective axes used. For example, in an embodiment using only a single axis, there will be a single stream of recorded states for that axis. In contrast, in an embodiment using two or more single axes, there will be separate streams of recorded states for each of those single axes. As another example, in an embodiment using one or more collective axes, there will be separate streams of recorded states for each of the collective axes.

[0278] While in one of the exemplary embodiments the flow proceeds from reference 1204 to 1206 (reference 1206 is performed by the WED), in an alternative of the exemplary embodiments the flow proceeds from reference 1206 to references 1212-1216 (reference 1212 is performed by the WED, and references 1214-1216 are performed in another electronic device coupled to the wearable electronic device (e.g., a server (containing hardware and software) / tablet computer / smartphone containing an application (called an app)).

[0279] Reference 1206 describes deriving the time span in which the WED was not worn based on the state recorded for consecutive time periods. In one embodiment, each of the time spans includes at least a threshold consecutive number of consecutive time periods in an unworn state. In one embodiment, reference 1206 is performed in real time and includes detecting at least a threshold consecutive number of consecutive time periods in an unworn state to derive the start of one of the time spans (allowing execution of reference 1108). Further reference herein... Figure 14A and 15A -B describes an exemplary embodiment of 1206. Although in one embodiment of the invention each of the time spans must be a minimum time length (e.g., see [reference]). Figure 15A -B) and is of variable time length, but alternative embodiments implement a fixed-length time span.

[0280] Reference 1212 indicates that data representing states recorded over a continuous time period is transmitted to another electronic device. While in one embodiment the data representing the state includes both time and the recorded state, in an alternative embodiment the time and the recorded state are compressed to form data representing the state as described above.

[0281] Reference 1214 indicates the receipt of data representing the state recorded for a continuous time period at another electronic device. Various techniques can be used to implement the communications represented by references 1212-1214, including wired / wireless and via one or more networks (including the Internet).

[0282] Reference 1215 describes the time span during which the WED was not worn, derived from data representing the state recorded over a continuous time period. Reference 1215 can be performed in a manner similar to Reference 1206, or in a more advanced manner (where another electronic device has greater processing power and storage capacity).

[0283] Optionally, reference 1216 describes transmitting an instruction from another electronic device to the WED after determining that the WED is currently in a time span during which the WED is not being worn. In one embodiment of the invention, the transmission of reference 1216 is utilized by the WED to perform reference 1108.

[0284] Operation of automatic non-wearable state detection

[0285] Figure 13 This describes an embodiment of the invention regarding the operation of non-wearable state detection using accelerated metrology. Figure 13 The task frames and blocks can be implemented within the wearable electronic device or distributed between the wearable electronic device and one or more other electronic devices coupled to the wearable electronic device, wherein the one or more other electronic devices may be, for example, electronic devices used to implement blocks 1322 / 1350 (e.g., a server (including hardware and software) / tablet computer / smartphone containing an application (called an app)). Task frames 1-4 illustrate the order in which the components shown by blocks 1312-1370 operate according to one embodiment of the invention. It should be understood that although the above discussion refers to accelerometers and acceleration data, other embodiments may utilize other types of motion sensors and motion data.

[0286] At task frame 1, accelerometer 1312 generates acceleration data for one or more axes. For example, acceleration may be sampled at 20Hz for one axis, with one sample every 0.05 seconds for that axis. When accelerometer 1312 generates acceleration data for only one axis, the data is forwarded to the first state recorder 1314, and the second state recorder 1315 and the third state recorder 1316 are not used (or even present). When acceleration data is generated for multiple axes and the acceleration data for said multiple axes will be used for automatic non-wearable state detection, one or more other state recorders (e.g., the second recorder 1315 and the third state recorder 1316) may be used.

[0287] At task frame 2, one or more state recorders record a time period as either a worn or unworn state based on whether acceleration data for a given axis satisfies a condition relative to a threshold acceleration for each consecutive time period. In one embodiment, each state recorder operates in a manner similar to that described with respect to reference 1204. That is, each recorder can compare motion data for a given axis with an unworn profile. As discussed above, a state recorder can record time periods based on acceleration data for a single axis or multiple axes.

[0288] As discussed above, the threshold acceleration can have different values ​​for different state recorders. For example, for non-wearable recorders, the threshold acceleration for common orientation A (e.g., as regarding...) can vary. Figure 10A The threshold acceleration (the Y-axis discussed) can be higher than that for another common orientation B (e.g., as discussed in this paper regarding...). Figure 10C The threshold acceleration (of the X-axis) discussed is such that, for example, common orientation A causes the Y-axis to extend closer to parallel to gravity than common orientation B causes the X-axis to extend parallel to gravity. In this case, the state recorder corresponding to common orientation A may have a higher threshold acceleration compared to the state recorder corresponding to common orientation B.

[0289] Subsequently, in task frame 3, the non-wearable time span classifier 1322 derives the time span during which the wearable electronic device was not worn based on data representing states recorded for continuous time periods. (The preceding text, for example, is related to...) Figure 12 Refer to 1206 and 1215 for discussions of various methods used to derive the aforementioned time span. Note that the threshold number of consecutive time periods recorded as non-wearable states for a single axis can be determined as follows: Figure 15A The differences will vary depending on the state recorder described in -B.

[0290] It should be noted that the state recorder and the non-wearable time span classifier utilize various thresholds (e.g., threshold acceleration and threshold number of consecutive time periods recorded as non-wearable states), and these thresholds may vary for different common orientations of wearable electronics when not worn, as may be specified by different non-wearable profiles.

[0291] Optionally at task frame 4A, based on a time span, the sleep tracker 1350 determines time blocks (which may not overlap and / or be consecutive), during which the user is in one of a plurality of states, said states including a waking state and a sleeping state, said consecutive time blocks having different states among said plurality of states. Task frame 4A may be implemented in various ways relative to those described with reference to 1106.

[0292] Also optionally at task frame 4B, the WED operation regulator 1370 regulates one or more of the following: WED power consumption, sensor data storage, data transmission / reception, and firmware upgrade scheduling. Task frame 4B can be implemented in various ways relative to those described with reference to 1108.

[0293] Figure 14A According to one embodiment, the recording of a time period as either a wearable or non-wearable state is based on acceleration data measured along the accelerometer axis exceeding a threshold acceleration for a consecutive time period. In each time period 1402-1406, several acceleration data samples are generated for the axis, and each acceleration sample is compared to a threshold acceleration used for recording a non-wearable state (refer to 1420). In this example, the time period is recorded as a non-wearable state when the values ​​of all samples in a time period exceed the threshold acceleration. For illustration, the accelerations sampled at references 1432 and 1436 all exceed the threshold acceleration, and therefore time periods 1402 and 1406 are recorded as non-wearable periods. In contrast, one value of the acceleration data in time period 1404 is below the threshold acceleration, and time period 1404 is recorded as a wearable state. This implementation of recording a non-wearable state only when all acceleration data exceeds the threshold can be used to reduce incorrect detection (false alarms) of non-wearable states.

[0294] Figure 14B This illustration describes how, according to one embodiment of the invention, a time span in which the wearable electronic device is not worn is derived based on states recorded for consecutive time periods. The figures illustrate the recorded states for the time periods, where each time period is recorded as either a worn (black block) or non-wearable (white block) state. The non-wearable time span comprises consecutive time periods in which the wearable electronic device is recorded as non-wearable. As illustrated, the time span at reference 1422 ends when a single time period of the worn state occurs. Note that for the non-wearable time span to be detected, the time span must be sufficiently long (exceeding a threshold for consecutive time periods). If a set of consecutive time periods recorded as non-wearable is interrupted before reaching the threshold, then the time span is derived as a worn state.

[0295] The threshold time period may be selected, for example, from a range of 5 minutes to 2 hours. Furthermore, the threshold time period may differ for different coordinators: for example, for the first coordinator (e.g., ... Figure 10A The threshold time is selected from a range of 5 minutes to 120 minutes (e.g., 30 minutes), while for a co-oriented second party (e.g., Figure 10B The time is 10 to 180 minutes (e.g., 60 minutes).

[0296] Figure 15AThis invention describes the detection of a non-wearable state for a time span relative to a first axis, according to one embodiment of the invention. In this embodiment, the time span in the non-wearable state may not have a fixed length and extends as long as no wearing state time period is detected. It is advantageous to be able to detect the time span in which the wearable electronics are not worn, as this information can be used, for example, to adjust the operation of the wearable electronics as discussed above herein. The detection, as described with reference to 1512, is based on the detection of at least a first threshold consecutive number of consecutive time periods recorded as non-wearable for the first axis. The first axis is based on the fact that the wearable electronics are typically placed in a certain orientation when not worn, and the first threshold consecutive number is based on the probability that the wearable electronics will remain in the orientation for a first threshold amount of time when worn.

[0297] Figure 15B This invention describes the detection of a non-wearable state for a second axis time span. As described with reference to 1522, the detection is based on at least a second threshold consecutive number of consecutive time periods recorded as non-wearable for the second axis. The second axis is based on the fact that the wearable electronics are typically placed in a certain orientation when not worn, and the second threshold consecutive number is based on the probability that the wearable electronics will remain in said orientation for a second threshold amount of time when worn. That is, the detection of a non-wearable state for different axis time spans can be based on different thresholds for consecutive number of time periods.

[0298] Exemplary device for implementing automatic detection of non-wearable state

[0299] As previously described, while in some embodiments the operations are implemented in wearable electronic devices, alternative embodiments distribute the differences in the operations across different electronic devices. Figure 16 (Illustrate an instance of this type of distribution). Figure 16 This is a block diagram illustrating a wearable electronic device and an electronic device that implement the operations disclosed in an embodiment of the present invention. Wearable electronic device (WED) 1602 and electronic device 1600 are similar to WED 902 and electronic device 900, respectively, and the same or similar references indicate elements or components having the same or similar functionality.

[0300] WED 1602 includes a motion sensor 1312 to generate motion data, such as acceleration data. It also has a non-transitory machine-readable storage medium 1618 containing one or more state recorders 1620 as discussed above herein. The non-transitory machine-readable storage medium 1618 may also include a non-wearable time span classifier 1322 to classify when a time span is derived in a non-wearable state. Similarly, the non-transitory machine-readable storage medium 1618 may also include a sleep tracker 1350 and a WED operation regulator 1370, associated with operations performed within the WED.

[0301] Electronic device 1600 has a non-transitory machine-readable storage medium 1648, which optionally includes a non-wearable time span classifier 1622 to classify when a derived time span is performed at electronic device 1600 in a non-wearable state; and it includes a sleep tracker 1650 to track when a determination of a user state, including wakefulness and sleep states, is performed at electronic device 1600. In one embodiment, the sleep tracker 1650 is... Figure 9 The sleep tracking module 950.

[0302] When executed by processor 1652, NTSC 1622 causes electronic device 1600 to perform the corresponding operations described above. Electronic device 1600 may contain virtual machines (VMs) 1662A to 1662R, each of which can execute a software instance of NTSC 1650. Super monitor 1654 may present a virtual operating platform for virtual machines 1662A to 1662R.

[0303] While the flowcharts in the figures above illustrate a particular order of operations performed by certain embodiments, it should be understood that this order is exemplary (e.g., alternative embodiments may perform the operations in a different order, combine particular operations, overlap particular operations, etc.).

[0304] Although the invention has been described with reference to several embodiments, those skilled in the art will recognize that the invention is not limited to the described embodiments and can be practiced with modifications and alterations within the spirit and scope of the appended claims. Therefore, the description should be considered illustrative rather than restrictive.

[0305] Alternative embodiments

[0306] Many specific details have been set forth herein. However, it should be understood that embodiments can be practiced without these specific details. In other instances, well-known circuits, structures, and techniques have not been shown in detail so as not to obscure the understanding of this description. However, those skilled in the art will understand that the invention can be practiced without these specific details. Those of ordinary skill in the art will be able to implement appropriate functionality without improper experimentation using the included description.

[0307] The embodiments described in this specification using references to "an embodiment," "an example," "an exemplary embodiment," etc., may include a specific feature, structure, or characteristic, but each embodiment may not necessarily include that specific feature, structure, or characteristic. Furthermore, these phrases do not necessarily refer to the same embodiment. Additionally, when a specific feature, structure, or characteristic is described in connection with an embodiment, it should be acknowledged that implementing that feature, structure, or characteristic in conjunction with other embodiments, whether explicitly described or not, is within the knowledge of those skilled in the art.

[0308] Parenthesized text and blocks with dashed boundaries (e.g., large dashes, small dashes, dotted-dash lines, and dots) may be used herein to illustrate optional operations of adding additional features to embodiments. However, this notation should not be construed as meaning that these are merely optional or choice operations, and / or that blocks with solid boundaries in some embodiments are not optional.

[0309] The terms “coupled” and “connected”, along with their derivatives, may be used in this description and the appended claims. It should be understood that these terms are not intended to be synonyms. “Coupled” is used to indicate the cooperation or interaction of two or more elements (which may or may not be in direct physical or electrical contact with each other). “Connected” is used to indicate the establishment of communication between two or more coupled elements. As used herein, “set” refers to any positive integer number of items containing one item.

[0310] The operations in the flowchart have been described with reference to exemplary embodiments in the other figures. However, it should be understood that the operations in the flowchart can be performed by embodiments other than those discussed with reference to the other figures, and the embodiments discussed with reference to these other figures can perform operations different from those discussed with reference to the flowchart.

[0311] The following examples, rather than limitations, provide some alternative embodiments.

[0312] Example 31. A device for automatically detecting the sleep cycles of a user of a wearable electronic device, the device comprising: a group of one or more processors; a non-transitory machine-readable storage medium coupled to the group of one or more processors and storing instructions therein, the instructions, when executed by the group of one or more processors, causing the device to: obtain a set of features for one or more time periods from motion data obtained from or derived from the group of one or more motion sensors; classify the one or more time periods into one of a plurality of states of the user based on the set of features determined for the one or more time periods, wherein the state indicates the relative degree of movement of the user; and derive time blocks covering the time periods, during which the user is in one of a plurality of states, wherein the states include a waking state and a sleeping state.

[0313] Example 32. The device according to Example 31, wherein the classification is performed using a machine learning classifier.

[0314] Example 33. The device according to Example 31, wherein the plurality of states further includes a non-wearable state, during which the user is not wearing the wearable electronic device.

[0315] Example 34. The device according to Example 31, wherein the device is one of the wearable electronic device and a secondary electronic device coupled to the wearable electronic device.

[0316] Example 35. The device according to Example 33, wherein the combination of the quantification of the distribution of the motion data comprises a combination of statistical measures of the motion data along each of the three axes.

[0317] Example 36. According to the device of Example 31, the classification of each of the time periods is further based on photoplethysmography (PPG) data from a time window including the time period, and wherein the PPG data is generated by a photoplethysmography sensor in the wearable electronic device.

[0318] Example 37. The device according to Example 36, wherein at least one of the following of the user is calculated using the PPG data: heart rate data; heart rate variability data; and respiratory data.

[0319] Example 38. The device according to Example 31, wherein the instructions, when executed by the group of processors, further cause the device to: after deducing that the user is asleep, determine another set of one or more statistical features for the time period during which the user is asleep, wherein at least one characterizes the distribution of the user's movement; and classify each of the time periods into multiple sleep stages of the user based on the set of statistical features determined for the time of interest, wherein the sleep stages include a rapid eye movement (REM) stage and multiple non-REM stages.

[0320] Example 39. According to the device of Example 31, the classification of each of the time periods is further based on data generated by a set of additional sensors in the wearable electronics, wherein the set includes one or more of the following: a temperature sensor; an ambient light sensor; a skin conductance sensor; a capacitive sensor; a humidity sensor; and a sound sensor.

[0321] Example 40. The device according to Example 31, wherein the set of motion sensors includes an accelerometer.

[0322] Example 41. A method comprising: obtaining a set of features for one or more time periods from motion data obtained from or derived from a set of one or more motion sensors; classifying the one or more time periods into one of a plurality of states of the user based on the set of features determined for the one or more time periods, wherein the state indicates the relative degree of movement of the user; and deriving time blocks covering the time periods, during which the user is in one of a plurality of states, wherein the states include an awake state and a asleep state.

[0323] Example 42. The method according to Example 41, wherein the classification is performed using a machine learning classifier.

[0324] Example 43. According to the method of Example 41, the plurality of states further includes a non-wearable state, during which the user is not wearing the wearable electronic device.

[0325] Example 44. According to the method of Example 41, the device is one of the wearable electronic device and a secondary electronic device coupled to the wearable electronic device.

[0326] Example 45. According to the method of Example 43, the combination of the quantification of the distribution of the motion data includes a combination of statistical measures of the motion data along each of the three axes.

[0327] Example 46. According to the method of Example 41, the classification of each of the time periods is further based on photoplethysmography (PPG) data from a time window including the time period, and the PPG data is generated by a photoplethysmography sensor in the wearable electronic device.

[0328] Example 47. The method according to Example 46, wherein at least one of the following of the user is calculated using the PPG data: heart rate data; heart rate variability data; and respiratory data.

[0329] Example 48. According to the method of Example 41, wherein the instructions, when executed by the group processor, further cause the device to: after deducing that the user is asleep, determine another set of one or more statistical features for the time period during which the user is asleep, wherein at least one characterizes the distribution of the user's movement; and classify each of the time periods into multiple sleep stages of the user based on the set of statistical features determined for the time of interest, wherein the sleep stages include a rapid eye movement (REM) stage and multiple non-REM stages.

[0330] Example 49. According to the method of Example 41, the classification of each of the time periods is further based on data generated by a set of additional sensors in the wearable electronics, wherein the set includes one or more of the following: a temperature sensor; an ambient light sensor; a skin conductance sensor; a capacitive sensor; a humidity sensor; and a sound sensor.

[0331] Example 50. The method according to Example 41, wherein the set of motion sensors includes an accelerometer.

[0332] Example 51. A computer-readable storage device comprising instructions that, when executed by one or more processors, cause the one or more processors to: obtain a set of features for one or more time periods from motion data obtained from or derived from a set of one or more motion sensors; classify the one or more time periods into one of a plurality of states of a user based on the set of features determined for the one or more time periods, wherein the state indicates the relative degree of movement of the user; and derive time blocks covering the time periods, during which the user is in one of a plurality of states, wherein the states include an awake state and a asleep state.

[0333] Example 52. A computer-readable storage device according to Example 51, wherein the classification is performed using a machine learning classifier.

[0334] Example 53. The computer-readable storage device according to Example 51, wherein the plurality of states further includes a non-wearable state, during which the user is not wearing the wearable electronic device.

[0335] Example 54. A computer-readable storage device according to Example 51, wherein the device is one of the wearable electronic device and a secondary electronic device coupled to the wearable electronic device.

[0336] Example 55. A computer-readable storage device according to Example 53, wherein the combination of the quantization of the distribution of the motion data comprises a combination of statistical measures of the motion data along each of the three axes.

[0337] Example 56. According to the computer-readable storage device of Example 51, the classification of each of the time periods is further based on photoplethysmography (PPG) data from a time window including the time period, and wherein the PPG data is generated by a photoplethysmography sensor in the wearable electronic device.

[0338] Example 57. A computer-readable storage device according to Example 56, wherein at least one of the following of the user is calculated using the PPG data: heart rate data; heart rate variability data; and respiratory data.

[0339] Example 58. A computer-readable storage device according to Example 51, wherein the instructions, when executed by the set of processors, further cause the device to: after deducing that the user is asleep, determine another set of one or more statistical features for the time period during which the user is asleep, wherein at least one characterizes the distribution of the user's movement; and classify each of the time periods into multiple sleep stages of the user based on the set of statistical features determined for the time of interest, wherein the sleep stages include a rapid eye movement (REM) stage and multiple non-REM stages.

[0340] Example 59. According to the computer-readable storage device of Example 51, the classification of each of the time periods is further based on data generated by a set of additional sensors in the wearable electronics, wherein the set includes one or more of the following: a temperature sensor; an ambient light sensor; a skin conductance sensor; a capacitive sensor; a humidity sensor; and a sound sensor.

[0341] Example 60. A computer-readable storage device according to Example 51, wherein the group of motion sensors includes an accelerometer.

[0342] Example 61. A wearable electronic device to be worn by a user, the wearable electronic device comprising: a set of sensors for detecting the user's physiological data or environmental data; a set of one or more processors coupled to the set of sensors; and a non-transitory machine-readable storage medium coupled to the set of one or more processors and storing instructions therein, the instructions, when executed by the set of one or more processors, causing the wearable electronic device to: reduce power consumption from at least one sensor of the set of sensors based on detecting that the user's state, as tracked by the wearable electronic device, has transitioned to a sleep state; and reverse the reduction in power consumption of the at least one sensor based on detecting that the user's state, as tracked by the wearable electronic device, has transitioned out of the sleep state.

[0343] Example 62. The wearable electronic device according to Example 61, wherein the group of sensors includes a photoplethysmography sensor to generate photoplethysmography (PPG) data of the user, wherein the photoplethysmography sensor includes a light source and a photodetector.

[0344] Example 63. The wearable electronic device according to Example 62, wherein the reduction in power consumption includes at least one of the following: a reduction in the sampling rate of the photoplethysmography sensor, a reduction in the sensitivity of the photoplethysmography sensor, and a reduction in the power level of the light source.

[0345] Example 64. The wearable electronic device according to Example 62, wherein the instructions, when executed by the group processor, also cause the wearable electronic device to: after the reduction, increase the power consumption of the photoplethysmography sensor to generate additional PPG data for sleep stage detection.

[0346] Example 65. In the wearable electronic device according to Example 64, the increase in the power consumption of the photoplethysmography sensor includes at least one of the following: an increase in the sampling rate of the photoplethysmography sensor, an increase in the sensitivity of the photoplethysmography sensor, and an increase in the power level of the light source.

[0347] Example 66. The wearable electronic device according to Example 61, wherein the group of sensors includes motion sensors to generate motion data of the user.

[0348] Example 67. The wearable electronic device according to Example 66, wherein the reduction in power consumption includes at least one of the following: entering a low-precision state of the motion sensor, a reduction in the sensitivity of the motion sensor, and a reduction in the sampling rate of the motion sensor.

[0349] Example 68. The wearable electronic device according to Example 66, wherein the motion sensor is an accelerometer.

[0350] Example 69. The wearable electronic device according to Example 66, wherein the instructions, when executed by the group processor, also cause the wearable electronic device to: periodically and temporarily increase the power consumption of the motion sensor after the reduction and before the reversal to generate additional motion data for sleep stage detection.

[0351] Example 70. In the wearable electronic device according to Example 69, the increase in the power consumption of the motion sensor includes at least one of the following: entering a high-precision state of the motion sensor, an increase in the sensitivity of the motion sensor, and an increase in the sampling rate of the motion sensor.

[0352] Example 71. A wearable electronic device to be worn by a user, the wearable electronic device comprising: a set of sensors including at least one of the following: a photoplethysmography (PPG) sensor for generating PPG data of the user, a motion sensor for generating motion data of the user, wherein the PPG sensor includes a light source and a photodetector; a set of one or more processors coupled to the PPG sensor; a non-transitory machine-readable storage medium coupled to the set of one or more processors and storing instructions therein, the instructions causing the wearable electronic device, when executed by the set of one or more processors, to: determine that the user is asleep based on the motion data generated by the motion sensor; increase the power consumption of at least one of the PPG sensor and the motion sensor in response to determining that the user is asleep; determine that the user is awake based on the motion data generated by the motion sensor; and reverse the increase in the power consumption of the at least one of the PPG sensor and the motion sensor in response to determining that the user is awake.

[0353] Example 72. The wearable electronic device according to Example 71, wherein the increase in power consumption results in an increase in PPG data generated by the PPG sensor, wherein the instructions, when executed by the group one or more processors, also cause the wearable electronic device to calculate at least one of the following: a set of heart rates of the user, a set of heart rate variability of the user, and a set of respiratory rates of the user.

[0354] Example 73. The wearable electronic device according to Example 71, wherein the additional data is motion data for generating motion measures by a quantized combination of motion data distributions along each of the three axes at time intervals, wherein each of the motion measures is a single numerical value.

[0355] Example 74. The wearable electronic device according to Example 71, wherein the increase in power consumption includes at least one of the following: an increase in the sampling rate of the photoplethysmography sensor, an increase in the sensitivity of the photoplethysmography sensor, and an increase in the power level of the light source.

[0356] Example 75. The wearable electronic device according to Example 71, wherein the increase in power consumption includes at least one of the following: entering a high-precision state of the motion sensor, an increase in the sensitivity of the motion sensor, and an increase in the sampling rate of the motion sensor.

[0357] Example 76. The wearable electronic device according to Example 71, wherein the light source is a light-emitting diode (LED).

[0358] Example 78. A method for managing the power consumption of a wearable electronic device, wherein the wearable electronic device includes a set of sensors, the method comprising: reducing power consumption from at least one sensor of the set of sensors based on detecting that the state of the user, as tracked by the wearable electronic device, has transitioned to a sleeping state; and reversing the reduction in power consumption of the at least one sensor based on detecting that the state of the user, as tracked by the wearable electronic device, has transitioned out of the sleeping state.

[0359] Example 79. The wearable electronic device according to Example 78, wherein the group of sensors includes a photoplethysmography sensor to generate photoplethysmography (PPG) data of the user, wherein the photoplethysmography sensor includes a light source and a photodetector.

[0360] Example 80. The wearable electronic device according to Example 79, wherein the reduction in power consumption includes at least one of the following: a reduction in the sampling rate of the photoplethysmography sensor, a reduction in the sensitivity of the photoplethysmography sensor, and a reduction in the power level of the light source.

[0361] Example 81. The wearable electronic device according to Example 79, wherein the instructions, when executed by the group processor, also cause the wearable electronic device to: periodically and temporarily increase the power consumption of the photoplethysmography sensor after the reduction and before the reversal to generate additional PPG data for sleep stage detection.

[0362] Example 82. In the wearable electronic device according to Example 81, the increase in the power consumption of the photoplethysmography sensor includes at least one of the following: an increase in the sampling rate of the photoplethysmography sensor, an increase in the sensitivity of the photoplethysmography sensor, and an increase in the power level of the light source.

[0363] Example 83. The wearable electronic device according to Example 78, wherein the group of sensors includes motion sensors to generate motion data of the user.

[0364] Example 84. The wearable electronic device according to Example 83, wherein the reduction in power consumption includes at least one of the following: entering a low-precision state of the motion sensor, a reduction in the sensitivity of the motion sensor, and a reduction in the sampling rate of the motion sensor.

[0365] Example 85. The wearable electronic device according to Example 83, wherein the instructions, when executed by the group processor, also cause the wearable electronic device to: periodically and temporarily increase the power consumption of the motion sensor after the reduction and before the reversal to generate additional motion data for sleep stage detection.

[0366] Example 86. In the wearable electronic device according to Example 85, the increase in the power consumption of the motion sensor includes at least one of the following: entering a high-precision state of the motion sensor, an increase in the sensitivity of the motion sensor, and an increase in the sampling rate of the motion sensor.

[0367] Example 87. A method for managing the power consumption of a wearable electronic device, wherein the wearable electronic device includes a set of sensors, including at least one of a photoplethysmography (PPG) sensor for generating photoplethysmography (PPG) data of the user and a motion sensor for generating motion data of the user, wherein the PPG sensor includes a light source and a photodetector; the method includes: increasing the power consumption of the at least one of the PPG sensor and the motion sensor in response to a state transition of the user as tracked by the wearable electronic device to a sleep state, wherein the increase in power consumption provides additional data for sleep stage detection; and reversing the increase in power consumption of the at least one of the PPG sensor and the motion sensor in response to a state transition of the user as tracked by the wearable electronic device to leave the sleep state.

[0368] Example 88. According to the method of Example 85, wherein the additional data is PPG data for calculating at least one of the following: a set of heart rates of the user, a set of heart rate variability of the user, and a set of respiratory rates of the user.

[0369] Example 89. According to the method of Example 85, wherein the additional data is motion data for generating motion measures by a quantized combination of the distribution of motion data along each of the three axes at time intervals, and wherein each of the motion measures is a single numerical value.

[0370] Example 90. According to the method of Example 85, the increase in power consumption includes at least one of the following: an increase in the sampling rate of the photoplethysmography sensor, an increase in the sensitivity of the photoplethysmography sensor, and an increase in the power level of the light source.

[0371] Example 91. A wearable electronic device includes: a group of one or more motion sensors for generating motion data; a group of one or more processors; and a non-transitory machine-readable storage medium coupled to the motion sensors and the group of one or more processors, the non-transitory machine-readable storage medium storing instructions that, when executed by the group of processors, cause the group of processors to: automatically determine a period of time during which the wearable electronic device has not been worn based on a comparison of the motion data with a non-wear profile, the non-wear profile specifying a pattern of motion data indicating when the wearable electronic device has not been worn by the user; and store data that associates the period of time with the non-wearable state in the non-transitory machine-readable storage medium.

[0372] Example 92. The wearable electronic device according to Example 91, wherein the motion data comprises a plurality of motion data samples, and the instruction, when executed by the group processor, causes the device to automatically determine, for the time period, that the wearable electronic device is not worn based on causing the group processor to perform the following operation: determining that the number of motion data samples that fail to meet a motion measurement threshold is less than the threshold number.

[0373] Example 93. According to the device of Example 91, wherein the non-transitory machine-readable storage medium further stores additional data that associates an additional time period with the non-wearable state, the time period and the additional time period together representing a continuous time period, and wherein the instructions, when executed, also cause the group of processors to: derive a time span covering the continuous time period based on the data and the additional data that associates the continuous time period with the non-wearable state; and store the data that associates the time span with the non-wearable state in the non-transitory machine-readable storage medium.

[0374] Example 94. The device according to Example 93, wherein the derivation of the time span includes instructions that, when executed, cause the group processor to: detect from the time period and the additional time period associated with the non-wearable state that the continuous time period contains at least a threshold number of consecutive time periods.

[0375] Example 95. The device according to Example 91, wherein the instructions, when executed by the group of processors, also cause the wearable electronic device to automatically cause one or more of the following based on the non-wearable state for the time period: a reduction in power consumption of the wearable electronic device, an interruption in the storage of sensor data from one or more sensors of the wearable electronic device, data transmission to another electronic device, and receiving a firmware update from the other electronic device.

[0376] Example 96. The device according to Example 91, wherein the instructions, when executed by the group processor, also cause the group processor to automatically determine another time period during which the wearable electronics were not worn based on a subsequent comparison of the motion data with another non-wearable profile, the other non-wearable profile specifying different patterns of motion data indicating when the wearable electronics were not worn by the user.

[0377] Example 97. The device according to Example 91, wherein the pattern of motion data characterizes the orientation of the wearable electronics.

[0378] Example 98. The device according to Example 97, wherein the pattern of the motion data further characterizes the acceptable range of motion for the orientation.

[0379] Example 99. The device according to Example 97, wherein the pattern of motion data is characterized by a threshold force that conforms to the acceleration along one or more axes of gravity applied along said one or more axes, representing the orientation.

[0380] Example 100. The device according to Example 97, wherein the pattern of motion data is characterized by a threshold force that conforms to the acceleration along one or more axes of gravity applied along said one or more axes, representing the orientation.

[0381] Example 101. The device according to Example 100, wherein the threshold force of acceleration takes into account the determinable tilt of the wearable electronics.

[0382] Example 102. The device according to Example 99, wherein the one or more axes represent axes extending across the display.

[0383] Example 103. The device according to Example 99, wherein the one or more axes represent axes extending through the display.

[0384] Example 104. The device according to Example 91, wherein the instructions, when executed by the group processor, also cause the group processor to transmit data representing the association between the non-wearable state and the time period to another electronic device.

[0385] Example 105. The device according to Example 91, wherein the instructions, when executed by the group processor, further cause the group processor to: automatically determine subsequent time periods in which the wearable electronics are worn based on a subsequent comparison of the motion data with the non-wearable profile, the non-wearable profile specifying the pattern of motion data indicating when the wearable electronics are not worn by the user; and store data that associates the subsequent time periods with the wearing state in the non-transitory machine-readable storage medium.

[0386] Example 106. A method executed by one or more processors of a wearable electronic device, the method comprising: obtaining motion data generated by one or more motion sensors of the wearable electronic device; determining a period of time during which the wearable electronic device is not worn based on a comparison of the motion data with a non-wear profile, the non-wear profile specifying a pattern of motion data indicating when the wearable electronic device is not worn by the user; and storing data that associates the period of time with the non-wear state in a non-transitory machine-readable storage medium of the wearable electronic device.

[0387] Example 107. According to the method of Example 106, wherein the motion data comprises a plurality of motion data samples, and the determination that the wearable electronic device is not worn for the time period comprises: determining that the number of motion data samples that fail to meet the motion measurement threshold is less than the threshold number.

[0388] Example 108. The method according to Example 106 further includes: storing additional data that associates an additional time period with the non-wearable state in the non-transitory machine-readable storage medium, the time period and the additional time period together representing a continuous time period; deriving a time span covering the continuous time period based on the data and the additional data that associates the continuous time period with the non-wearable state; and storing the data that associates the time span with the non-wearable state in the non-transitory machine-readable storage medium.

[0389] Example 109. According to the method of Example 108, the derivation includes: detecting from the time period and the additional time period associated with the non-wearable state that the continuous time period contains at least a threshold number of consecutive time periods.

[0390] Example 110. The method according to Example 106 further includes: causing one or more of the following based on the non-wearable state for the time period: reduced power consumption of the wearable electronics, interruption of storage of sensor data from one or more sensors of the wearable electronics, data transmission to another electronic device, and receiving firmware updates from the other electronic device.

[0391] Example 111. The method according to Example 106 further includes: automatically determining another time period during which the wearable electronics were not worn based on a subsequent comparison of the motion data with another non-wearable profile, the other non-wearable profile specifying different patterns of motion data indicating when the wearable electronics were not worn by the user.

[0392] Example 112. The method according to Example 106, wherein the pattern of the motion data characterizes the orientation of the wearable electronics.

[0393] Example 113. According to the method of Example 112, the pattern of the motion data further characterizes the acceptable range of motion for the orientation.

[0394] Example 114. According to the method of Example 112, the pattern of motion data is characterized by a threshold force that conforms to the acceleration along one or more axes of gravity applied along said one or more axes.

[0395] Example 115. According to the method of Example 112, the pattern of motion data is characterized by a threshold force that conforms to the acceleration along one or more axes of gravity applied along said one or more axes.

[0396] Example 116. The method according to Example 115, wherein the threshold force of acceleration takes into account the determinable tilt of the wearable electronics.

[0397] Example 117. The method according to Example 114, wherein the one or more axes represent axes extending across the display.

[0398] Example 118. The method according to Example 114, wherein the one or more axes represent axes extending through the display.

[0399] Example 119. The method according to Example 106 further includes: transmitting data representing the association between the non-wearable state and the time period to another electronic device.

[0400] Example 120. The method according to Example 106 further includes: automatically determining subsequent time periods in which the wearable electronics are worn based on a subsequent comparison of the motion data with the non-wearable profile, the non-wearable profile specifying the pattern of motion data indicating when the wearable electronics are not worn by the user; and storing data that associates the subsequent time periods with the wearing state in the non-transitory machine-readable storage medium.

Claims

1. A wearable electronic device comprising: one or more motion sensors configured to generate motion data; one or more processors; and a non-transitory machine-readable storage medium coupled to the one or more motion sensors and the one or more processors, the non-transitory machine-readable storage medium having stored therein instructions that, when executed by the one or more processors, cause the wearable electronic device to: automatically determine a time period in which the wearable electronic device was not worn by a user of the wearable electronic device in response to: i) for a duration of a time period, the motion data being greater than a first threshold along a first axis of the wearable electronic device; and ii) the time period being greater than a threshold amount of time, and store data in the non-transitory machine-readable storage medium that associates the time period with a non-worn state, wherein the non-transitory machine-readable storage medium further stores additional data that associates additional time periods with the non-worn state, the time period and the additional time periods together representing a contiguous time period, and wherein the instructions, when executed, further cause the wearable electronic device to: derive a time span covering the contiguous time period based on the additional data and the data that associates the contiguous time period with the non-worn state; and store data in the non-transitory machine-readable storage medium that associates the time span with the non-worn state.

2. The wearable electronic device of claim 1, wherein the motion data comprises motion data samples, and the instructions, when executed by the one or more processors, cause the wearable electronic device to automatically determine that the wearable electronic device was not worn by the user for the time period based on causing the wearable electronic to: determine that a number of motion data samples that fail to satisfy a motion metric threshold is below a threshold number.

3. The wearable electronic device of claim 1, wherein the instructions, when executed by the one or more processors, cause the derivation of the time span to cause the wearable electronic device to detect that the contiguous time period comprises at least a threshold consecutive number of time periods.

4. The wearable electronic device of claim 1, wherein the instructions, when executed by the one or more processors, further cause the wearable electronic device to: automatically cause one or more of a reduction in power consumption of the wearable electronic device, an interruption in storage of sensor data from one or more sensors of the wearable electronic device, a transfer of data to another electronic device, and a receipt of a firmware update from another electronic device based on the non-worn state for the time period.

5. The wearable electronic device of claim 1, wherein the instructions, when executed by the one or more processors, further cause the wearable electronic device to automatically determine another time period in which the wearable electronic device was not worn by the user in response to the motion data being greater than a second threshold along a second axis of the wearable electronic device, the second axis being different than the first axis. ​ 6. The wearable electronic device of claim 1, wherein the motion data being greater than the first threshold along the first axis indicates an orientation of the wearable electronic device.

7. The wearable electronic device of claim 6, wherein the motion data being greater than the first threshold along the first axis indicates that the orientation of the wearable electronic device has a tilt within a defined range.

8. The wearable electronic device of claim 6, wherein the first threshold relates to a gravitational force exerted along the first axis.

9. The wearable electronic device of claim 8, wherein the first axis comprises an axis extending across a display of the wearable electronic device.

10. The wearable electronic device of claim 8, wherein the first axis comprises an axis extending through a display of the wearable electronic device.

11. The wearable electronic device of claim 8, wherein the first threshold accounts for a determinable tilt of the wearable electronic device relative to a direction of gravity.

12. The wearable electronic device of claim 1, wherein the instructions, when executed by the one or more processors, further cause the wearable electronic device to: transmit data representing an association of the non-wearing state with the time period to another electronic device.

13. The wearable electronic device of claim 1, wherein the instructions, when executed by the one or more processors, further cause the wearable electronic device to: in response to a subsequent determination that the motion data along the first axis of the wearable electronic device is less than or equal to the first threshold, automatically determine a subsequent time period in which the wearable electronic device is worn by the user; and store data associating the subsequent time period with a wearing state in the non-transitory machine-readable storage medium.

14. A method performed by a set of one or more processors of a wearable electronic device, the method comprising: obtaining motion data generated by one or more motion sensors of the wearable electronic device; automatically determining, by one or more processors of the wearable electronic device, a time period in which the wearable electronic device is not worn by a user of the wearable electronic device in response to: i) the motion data being greater than a threshold along a first axis of the wearable electronic device for a duration of a time period; and ii) the time period being greater than a threshold amount of time; storing data associating the time period with a non-wearing state in a non-transitory machine-readable storage medium; storing additional data associating additional time periods with the non-wearing state in the non-transitory machine-readable storage medium, the time period and the additional time periods together representing consecutive time periods; deriving a time span covering the consecutive time periods based on the additional data and the data associating the consecutive time periods with the non-wearing state; and storing data associating the time span with the non-wearing state in the non-transitory machine-readable storage medium.

15. The method of claim 14, wherein the motion data comprises motion data samples, and determining that the wearable electronic device was not worn by the user for the time period comprises: determining that a number of motion data samples failing to satisfy a motion metric threshold is below a threshold number.

16. The method of claim 14, wherein the deriving comprises: detecting that the consecutive time periods include at least a threshold number of consecutive time periods.

17. The method of claim 14, further comprising: based on the non-wear state for the time period, automatically causing one or more of: a reduction in power consumption of the wearable electronic device, an interruption of storage of sensor data from one or more sensors of the wearable electronic device, a data transfer to another electronic device, and a receipt of a firmware update from another electronic device.

18. The method of claim 14, further comprising: in response to the motion data being greater than a second threshold along a second axis of the wearable electronic device, automatically determining another time period in which the wearable electronic device was not worn by the user, the second axis being different than the first axis.

19. The method of claim 14, wherein the motion data being greater than a first threshold along the first axis is indicative of an orientation of the wearable electronic device.

20. The method of claim 19, wherein the motion data being greater than the first threshold along the first axis is indicative of the orientation of the wearable electronic device having a tilt within a defined range.

21. The method of claim 19, wherein the first threshold relates to a gravitational force exerted along the first axis.

22. The method of claim 21, wherein the first axis comprises an axis extending across a display of the wearable electronic device.

23. The method of claim 21, wherein the first axis comprises an axis extending through a display of the wearable electronic device.

24. The method of claim 21, wherein the first threshold accounts for a determinable tilt of the wearable electronic device relative to a direction of gravity.

25. The method of claim 14, further comprising: transmitting data representative of an association of the non-wear state with the time period to another electronic device.

26. The method of claim 14, further comprising: in response to a subsequent determination that the motion data is less than or equal to a first threshold along a first axis of the wearable electronic device, determining a subsequent time period in which the wearable electronic device was worn by the user; and storing data in the non-transitory machine-readable storage medium that causes the subsequent time period to be associated with a wear state.

27. The method of claim 14, further comprising: in response to a determination that the motion data is less than or equal to a first threshold along a first axis of the wearable electronic device, determining a time period in which the wearable electronic device was worn by the user; and storing data in the non-transitory machine-readable storage medium that causes the time period to be associated with a wear state.

28. The method of claim 14, further comprising: in response to a determination that the motion data is greater than a first threshold along a first axis of the wearable electronic device, determining a time period in which the wearable electronic device was not worn by the user; and storing data in the non-transitory machine-readable storage medium that causes the time period to be associated with a non-wear state.

29. The method of claim 14, further comprising: in response to a determination that the motion data is greater than a first threshold along a first axis of the wearable electronic device, determining a time period in which the wearable electronic device was not worn by the user; and storing data in the non-transitory machine-readable storage medium that causes the time period to be associated with a non-wear state.

30. The method of claim 14, further comprising: in response to a determination that the motion data is less than or equal to a first threshold along a first axis of the wearable electronic device, determining a time period in which the wearable electronic device was worn by the user; and storing data in the non-transitory machine-readable storage medium that causes the time period to be associated with a wear state.

Images