JP6867448B2 - Devices and methods for classifying user activity and / or counting user steps - Google Patents

Description

Comprehensively, the present invention relates to devices and methods for classifying user activities and / or counting user steps.

In recent years, people have become more and more aware of their health and have begun running and walking as an affordable and effective way of exercising. There is also growing interest in self-tracking daily activities through wearable sensors. Therefore, there is a need for a step count device that can accurately record not only the number of steps taken but also the walking state of the wearer.

Traditional step counting devices using accelerometers (s) usually require the user to first position the device in a limited set of orientations. For example, the user must secure the device to his body so that the dedicated axis (ie, the direction of the device) is substantially and continuously aligned with the direction of gravity throughout the measurement period.

Existing devices are typically unable to detect changes corresponding to the user's different walks or distinguish between signals corresponding to the user's different walks, resulting in inaccurate counts of steps taken. It becomes a thing. In the existing step counting device, the operation noise received by the device also measures the incorrect foot movement, and the actual foot movement is overlooked.

Also, the forward and backward swing of the body part to which the device is attached is not the same for all foot movements, and existing devices are usually due to inconsistent signal peaks and valleys. Therefore, the peak cannot be accurately identified as a step count.

Embodiments of the present invention provide activity monitoring systems and methods that attempt to address at least one of the above problems.

According to the first aspect of the present invention, it is a device that classifies the activities of the user who wears the device.
Accelerometers that measure accelerometer data for multiple axes and
It ’s a processor,
Identify the most active one of the multiple axes based on accelerometer data
Classify user activity based on the signal amplitude of the most active axis accelerometer data and one or more thresholds.
The processor, which is configured to
The device is provided.

According to the second aspect of the present invention, the device counts the number of steps taken by the user wearing the device.
An accelerometer that measures accelerometer data for at least one axis,
It ’s a processor,
Apply the differential operator to the accelerometer data in the continuous processing window
Count the peak of the derivative of the accelerometer data in each processing window and
Remove overcounted peaks based on the time difference between the first peak in the current window and the last peak in the preprocessing window.
The processor, which is configured to
The device is provided.

According to the third aspect of the present invention, it is a method of classifying user activities.
Measuring accelerometer data for multiple axes and
Identifying the most active axis of multiple axes based on accelerometer data,
Classification of user activity based on the signal amplitude of the accelerometer data of the most active axis and one or more thresholds.
Methods are provided, including.

According to the fourth aspect of the present invention, it is a method of counting the number of steps taken by the user.
Measuring accelerometer data for at least one axis and
Applying differential operators to accelerometer data in continuous processing windows,
Counting the peak of the derivative of the accelerometer data in each processing window,
Removing overcounted peaks based on the time difference between the first peak in the current window and the last peak in the preprocessing window,
Methods are provided, including.

Embodiments of the present invention will be better understood and readily apparent to those skilled in the art from the following description, along with the drawings, merely as an example.

It is a figure which shows the process of activity classification by an example embodiment. It is a figure which shows the spike in the raw accelerometer data. FIG. 5 shows the accelerometer data of FIG. 2 with attenuated spikes after applying the smoothing process according to an exemplary embodiment. It is a figure which shows the example signal amplitude and the obtained characteristic of the accelerometer signal by an example embodiment. It is a figure which shows the process of classifying an activity using the threshold level by an example embodiment. It is a figure which shows the signal amplitude measured during the different activity by an example embodiment. It is a figure which shows the process of peak detection and step count according to an Example embodiment. It is a figure which shows the signal waveform of the wrist-worn device before and after the application of the Pan Tompkins derivation operator according to an example embodiment. It is a figure which shows the _{adaptive threshold value (Δ w} ) used for processing of the subsequent window (s) according to an example embodiment. It is a figure which shows the peak-to-peak distance between the peaks of two consecutive windows by an example embodiment. It is a figure which shows the process flow which performs activity classification and step count according to an Example Embodiment. It is a schematic diagram of an assembly including a wearable device and a communication device according to an exemplary embodiment. FIG. 5 is a block diagram of an assembly comprising a wearable device and a communication device according to an exemplary embodiment. It is a block diagram of the device which classifies the activity of the user by an example embodiment. It is a block diagram of the device which counts the number of steps of a user by one Example Embodiment. It is a flowchart which shows the method of classifying the activity of a user by an example embodiment. It is a flowchart which shows the method of counting the number of steps of a user by one Example Embodiment.

Embodiments of the present invention relate to wearable devices and methods for classifying user activities and wearable devices and methods for counting the number of steps taken by the user. In one non-limiting example embodiment, the device is worn on the user's wrist to acquire and process the user's acceleration information, classify the user's activity and / or count the number of steps taken by the user. It is configured to do.

The present specification also discloses a device that performs the operation of the above method, which device can exist inside and / or outside the wearable device in an exemplary embodiment. Such devices may be specially constructed for the intended purpose, or may include general purpose computers or other devices that are selectively activated or reconfigured by computer programs stored in the computer. The algorithms and indications presented herein are not inherently associated with any particular computer, nor are they associated with any other device. Various general purpose machines can be used with the programs as taught herein. Alternatively, it may be appropriate to build a more specialized device that performs the required method steps. The structure of a conventional general-purpose computer will be clarified from the following description. In addition, the specification also implicitly discloses computer programs in that it will be apparent to those skilled in the art that individual steps of the methods described herein can be performed by computer code. It is intended that this computer program is not limited to any particular programming language and its embodiments. It will be appreciated that various programming languages and their encodings can be used to implement the teachings of the disclosures contained herein. Furthermore, it is also intended that computer programs are not limited to any particular control flow. There are many other variants of computer programs that can use different control flows without departing from the spirit or scope of the invention.

Moreover, one or more of the steps in a computer program can be performed in parallel rather than sequentially. Such computer programs can be stored on any computer-readable medium. The computer-readable medium can include storage devices such as magnetic disks or optical disks, memory chips, or other storage devices suitable for interfacing with general purpose computers. Computer-readable media can also include hard-wired media such as those exemplified by Internet systems, or wireless media such as those exemplified by GSM® mobile phone systems. When a computer program is loaded and executed on such a general purpose computer, it effectively provides a device that performs the steps of the preferred method.

The present invention can also be implemented as a hardware module. More specifically, in the sense of hardware, a module is a functional hardware unit designed to be used with other components or modules. For example, the module can be implemented with discrete electronic components or can form part of a complete electronic circuit such as an application specific integrated circuit (ASIC). There are so many other possibilities. Those skilled in the art will appreciate that this system can also be implemented as a combination of hardware and software modules.

FIG. 1 shows a flowchart 100 showing a method of classifying user activities according to an exemplary embodiment. This method
In step 102, acquiring the user's acceleration signal through the wearable device
In step 104, dividing the acceleration signal data into a plurality of processing windows,
In each processing window
In step 106, smoothing the data using a moving average function
In step 108, the accelerometer data is converted into g value data, and
In step 110, identifying the most active axis of the acceleration signal and
In step 112, classifying the user's activity based on the acceleration signal from the most active axis,
including.

Details of the exemplary features that can be used separately and / or in combination with one or more of the other exemplary features are described below.

Processing window The size of the window is described herein as the duration of the window, by way of example only. Suitable accuracy for classifying activities and / or counting steps is not limited, but as an example, using a window size of about 0 (not including 0) seconds to 5 seconds depending on the specific accuracy requirement. It is known that it can be achieved by. In general, the longer the window size, the less chance of peaks being overcounted within the window transition area, and thus the higher the accuracy. The shorter the window, the better the transition between different activities that can occur within a very short time period.

Smoothing accelerometer data Raw accelerometer data is usually noisy. The stray spike 200 in the exemplary accelerometer data 202 is shown in FIG. The floating spike 200 can result in the erroneous axis being identified as the most active axis for subsequent processing. In an exemplary embodiment, smoothing is applied favorably to raw accelerometer data 202. Preferably, smoothing is achieved by applying a moving average operation to the raw accelerometer data 200.

FIG. 3 shows ^{accelerometer data 202 *} after smoothing according to one exemplary embodiment ^{, preferably with attenuated spikes 200 *} that do not affect the identification of the most active axis of subsequent processing.

Most active axis The most active axis in the illustrated embodiment is identified as an accelerometer axis having a maximum amplitude signal that represents the effect of gravity and the impact force of the wearer's foot on the ground. In other words, the most active axis in the exemplary embodiment is the accelerometer axis closest to the vertical direction, i.e., substantially perpendicular to the surface of the earth.

Advantageously, the most active axis can be more easily identified than any other axis, regardless of the position or orientation in which the device is used or worn by the user.

The most active axes in a preferred embodiment are the three smoothed x-axis, y-axis and z-axis (ie, regardless of the actual orientation with respect to the ground) within each window, referred to herein as moving averages. ) Is selected by comparing the data. The absolute value of each axis from each window is determined, and this absolute value is referred to herein as the absolute moving average. The average value of each axis based on this absolute value is then obtained, and this average value is referred to as an average absolute moving average in the present specification. From the comparison of these three average values, the axis with the maximum average value is identified as the most active axis of each window in the preferred embodiment.

Activity classification In the exemplary embodiment, the user's activity is classified into a light activity state, a moderate activity state, or a heavy activity state using the moving average data from the most active axis. To.

In one embodiment, the signal amplitude of the most active axis is the maximum of the most active axis data 406, for example, in a window 408 having a size of about 0 (not including 0) s-5s, as shown in FIG. It is calculated by subtracting the average g value 400 or the minimum (min) g value 402 from the max) g value 404.

FIG. 5 shows a flowchart 500 showing a process of classifying activities using threshold levels according to an exemplary embodiment. In the example where the device is wrist-worn by the user, experimental results (see FIG. 6) show that, for example, threshold levels of 0.1 g and 1 g can be used for activity classification.

In step 502, the signal amplitude (SA) of the most active axis is obtained. In step 504, when SA is smaller than the threshold value 1, for example, when SA <0.1 g, classification into a low activity state is performed (see step 506). In step 508, when SA is smaller than the threshold value 2, for example, when SA <1 g, classification into a medium active state is performed (see step 510). In other words, in one example of 0.1 g ≤ SA <1 g, the classification is done in a medium active state. When SA ≧ threshold value 2, for example, when SA ≧ 1 g, classification into a high activity state is performed (see step 512).

Examples of classification include:
Low activity-Standing up, waiting for traffic lights to turn blue or non-moving movements such as typing (eg on a desk) Medium activity-Slow walking, snappy walking High activity-Jogging, running

Most other activities involving physical movement, such as climbing stairs, cycling, swimming, playing football, are also examples of embodiments, for example.
It can be classified based on further division into multiple SA thresholds (above thresholds 1 and 2 in the above example) corresponding to each of these different activities, and / or the above three general classifications. Further processing of, in the exemplary embodiment,
Filtering and identification and / or signal matching techniques using frequency classification,
Can be done by.

Referring to FIG. 6, plots 600 and 602 are the SA of the most active axis calculated based on the maximum value minus the minimum value in the processing window, respectively, and the average value subtracted from the maximum value in the processing window. It shows the SA of the most active axis calculated based on the above. As can be seen from FIG. 6, it can be preferred for threshold-based classification that the SA is based on the maximum value in the processing window minus the average value. This is because the absolute values within each active region are reduced, resulting in a smoother plot with a narrower range of values.

FIG. 7 shows a flowchart 700 showing a process of counting the number of steps according to an exemplary embodiment. In step 702, accelerometer data is captured. In step 704, this data is divided into process windows. In step 706, lowpass filtering is applied to the data from the most active axis based on the classified activity. In step 708, the derivative of the filtered data is calculated using the Pantonpukins derivative operator. In step 710, the output of the derivative signal is normalized. In step 712, peak detection is performed to count the number of steps. In step 714, the adaptive threshold is determined based on the current process window and updated for use in the next process window. In step 716, the overcounted peak is deleted.

Details of the exemplary features that can be used separately and / or in combination with one or more of the other exemplary features are described below.

Low-pass filter based on activity type (step 706, Figure 7)
The cutoff frequency of the lowpass filter is preferably adjusted based on the activity type to eliminate operating noise in the correct frequency range. Choosing the correct cutoff frequency can advantageously improve the signal geometry to avoid false or undetected peaks during subsequent peak detection processing.

In an exemplary embodiment, the activity type can be entered by the user and can be identified using the activity classification methods and devices described above or by any other method or device.

ここで、ｎはデータ点を示し、ｎ＞４である。 Pantonpukins differential operator (step 708, FIG. 7)

Here, n indicates a data point, and n> 4.

The Pantonpukins derivative operator in equation (1) is used in a preferred embodiment to further improve the accuracy of peak detection by favorably expanding the gradient information and optimizing the signal state of peak detection. ..

During normal walking and running, for example, arm swings are usually synchronized with the opposite leg. For example, when the right arm swings forward, the left leg moves forward, and vice versa. For this reason, one acceleration peak is detected, for example, in a wrist-worn device for each footstep performed by the user.

Changes in acceleration may also depend on changes in hand waving and foot impact. During normal walking conditions, the impact force is usually not strong enough, for example most of the acceleration changes in wrist-worn devices are due to hand waving. Since the forward and backward swings of the hand are usually not the same for each foot movement, mismatched peaks and valleys in the signal waveform can make it difficult to implement an accurate peak detection algorithm, which is preferable. In embodiments, the Pantonpukins differential operator gives a more consistent pattern of peaks and valleys. 8a and 8b show the signal waveforms of the wrist-worn device before (curve 800) and after (curve 802) the Pantonpukins differential operator is applied, respectively, according to an exemplary embodiment.

Normalization (step 710, FIG. 7)
The normalization process in the exemplary embodiment is preferably such that the derivative signal is accurately scaled for peak detection processing. In one embodiment, the signal is normalized by first subtracting the minimum value of the signal from the signal and then dividing by the maximum value of the subtracted signal.

Peak detection (step 712, FIG. 7)
According to an exemplary embodiment, peaks are detected (ie, steps are counted) when the normalized derivative signal exceeds a threshold level. Since the data is normalized at this time (for example, min value = 0, max value = 1), the threshold value Δ1 of the _first processing window is, for example, arbitrary from 0 (not including 0) to 0.5. Can be set to a predetermined value of.

ここで、Δ_ｗ＝次のウィンドウｗの適応的閾値であり、ａ＝現在のウィンドウ（ｗ−１）内の最後のピーク値であり、ｂ＝この現在のウィンドウ（ｗ−１）内の最後の谷値であり、ｗ≧２である。 Determining and updating adaptive thresholds (step 714, FIG. 7)
As shown in FIG. 9, an adaptive threshold (Δ _w ) is determined and used in subsequent windows. This adaptive threshold can be set to any predetermined value of about 0 (not including 0)% to 50% of the difference between the last peak and valley values of the current window. In one example, the following equation is obtained.

Here, Δ _w = the adaptive threshold of the next window w, a = the last peak value in the current window (w-1), and b = the last in this current window (w-1). It is the valley value of, and w ≧ 2.

Removal of overcounted peaks (step 716, FIG. 7)
The inventors have found that Pantonpukins derivative operations can tend to produce the wrong first peak in subsequent windows, resulting in overcounting of peaks, i.e. steps. Has been done.

FIG. 10 shows a method of advantageously identifying such false peaks that are removed from the count in a preferred embodiment. The erroneous peak is identified by examining the distance d between the first peak 1000 of the current measurement window 1002 and the last peak 1004 of the previous window 1006.

The threshold can be set to 4 steps per second in one embodiment. For example, in a data point region with a sampling rate of 80 Hz, the distance threshold can be set to 20 data points based on 4 steps per second. If the peak-to-peak distance d is less than 20 data points, the current peak 1200 is identified as erroneous and removed from the step count.

FIG. 11 shows a flowchart showing the process flow 1100 of one embodiment. In this embodiment, both the activity classification technique and the activity-based step counting technique are integrated into a single wrist wearable device.

In step 1102, the user's acceleration data is captured by the device through a 3-axis accelerometer. In step 1104, the accelerometer signal is divided into processing windows of about 0 (not including 0) seconds to 5 seconds. In step 1106, the signals in each window are smoothed by performing a moving average function. In step 1108, the most active axis is identified by comparing the average absolute moving averages of each axis. The axis with the maximum value is the most active axis.

In steps 1110 and 1114, the activity in each window is classified by calculating the signal amplitude (maximum-average) of the moving average signal of the most active axis. Classification is based on thresholds of 0.1 g and 1 g in one non-limiting example. When an inactive state is identified (step 1110), the process moves to the next window because there are no steps to count (step 1112).

In the medium and high activity states (step 1114), step counting is performed. In steps 1116 and 1118, the cutoff frequency of the lowpass filter for each window is updated based on the identified activity and the lowpass filtering is applied accordingly.

Next, in step 1120, the Pantonpukins differential operator is applied. Then, in step 1122, the derivative signal is (in one non-limiting example, first subtract the minimum value of that signal from the signal and then divide by the maximum value of the subtracted signal). It is normalized and the signal amplitude is adjusted within each window, eg, in the range 0-1.

Next, in step 1124, peak detection is performed in which each detected peak contributes to the step count. In the first window after transitioning from a low activity state to a medium activity state or a high activity state, the threshold is set to 0 (not including 0) to 0.5 in one non-limiting example. In other words, the first window that transitions from the inactive state usually does not require step removal. For subsequent windows (in medium or high activity), overcounted peak detection and deletion is favorably performed until a low activity is detected for that window. In subsequent windows, adaptive thresholds are used, and in step 1126, in one non-limiting example, the height difference between the last peak and valley of the preceding window is about 0 (not including 0). It is set to% to 50%.

Then, in step 1128, the erroneous peaks resulting from the Pantonpukins operation are removed using, for example, a peak-to-peak distance (first peak in the current window vs. last peak in the previous window) threshold. The threshold in one non-limiting example is based on 4 steps per second. For example, in a data point region with a sampling rate of 80 Hz, the distance threshold is set to 20 data points. If the peak-to-peak distance is less than 20 data points, the current peak is identified as false and removed from the step count.

The process in the next window loops back and performs steps 1106-1128 as described above.

FIG. 12 shows an assembly 1200 with a wearable device in the form of a wristwatch 1201 according to an exemplary embodiment. It will be appreciated that in different embodiments, the device may be in any other form suitable for wearing on any part of the user's body, such as the user's arms, waist, hips or feet. The watch 1201 classifies the user's activity and / or calculates the number of steps taken by the user and outputs the result (s) to the telecommunications device of assembly 1200, such as the mobile phone 1202 or other portable electronic device. Alternatively, it wirelessly communicates with computing devices such as desktop computers, laptop computers, and tab computers.

FIG. 13 shows a schematic block diagram of an assembly 1300 comprising a wearable device 1301 according to an exemplary embodiment that classifies user activity and / or counts steps. The device 1301 includes a signal detection module 1302 such as an accelerometer or a gyroscope that acquires user acceleration information.

One non-limiting example of a preferred accelerometer that can be adapted for use in this device is the 3-axis accelerometer MMA8562FC available from Freescale Semiconductor, Inc. This accelerometer can offer the advantage of measuring acceleration in all three directions using a single package. Alternatively, several uniaxial accelerometers oriented to provide triaxial detection can be used in different embodiments.

The device 1301 also includes a data processing calculation module 1304 such as a processor configured to receive and process acceleration information from the signal detection module 1302. The device 1301 also includes a display unit 1306 that displays the result to the user of the device 1301. The device 1301 in this embodiment further comprises a wireless transmission module 1308 configured to wirelessly communicate with the telecommunications device 1310 in assembly 1300. The telecommunications device 1310 includes a wireless receiver module 1312 that receives a signal from the wearable device 1301 and a display unit 1314 that displays the result to the user of the telecommunications device 1310.

FIG. 14 shows a block diagram of a device that classifies the activities of a user wearing the device 1400 according to one embodiment. The device 1400 identifies the accelerometer 1402 that measures the accelerometer data of a plurality of axes, the most active axis among the plurality of axes based on the accelerometer data, and the signal amplitude and the signal amplitude of the accelerometer data of the most active axis. It comprises a processor 1404 configured to classify user activity based on one or more thresholds.

Processor 1404 can be configured to apply smoothing to accelerometer data before identifying the most active axis.

Processor 1404 can be configured to identify the most active axis by finding the axis with the maximum average absolute amplitude of the smoothed signal. This maximum average absolute amplitude is sometimes referred to as the maximum average absolute moving average signal value.

Processor 1404 can be configured to classify activity based on the maximum signal amplitude of accelerometer data for the most active axis.

Processor 1404 can be configured to classify activity based on the maximum signal amplitude of the accelerometer data for the most active axis minus the average signal amplitude.

Processor 1404 can be configured to identify the most active axis and classify user activity based on accelerometer data measured over a predetermined processing window.

The processing window can range from about 0 (not including 0) seconds to 5 seconds.

FIG. 15 shows a block diagram of a device 1500 that counts the number of steps taken by a user wearing the device according to one embodiment. The device 1500 applies the accelerometer 1502, which measures accelerometer data for at least one axis, and the differential operator to the accelerometer data in successive processing windows, and peaks the derivatives of the accelerometer data in each processing window. It includes a processor 1504 that is configured to count and remove overcounted peaks based on the time difference between the first peak in the current window and the last peak in the preprocessing window.

Processor 1504 is configured to count peaks based on a first threshold when the current processing window is the first processing window after transitioning from a low activity state to a medium activity state or a high activity state. be able to.

The processor can be further configured to update the adaptive threshold of the next window. The adaptive threshold of the next window can be based on the difference in signal amplitude of the last consecutive peak and valley pair in the current processing window. The adaptive threshold for the next window can be about 0 (not including 0)% to 50% of the difference in signal amplitude between the last consecutive peak and valley pair in the current processing window.

Processor 1504 is configured to remove a peak from the peak count in the current window if the time difference between the first peak in the current window and the last peak in the preprocessing window is less than the second threshold. can do. This second threshold can be about 1/4 second. Derivative operators can include Pantompukins derivative operators.

Devices 1400 and 1500 can be mounted on wearable devices.

Devices 1400 and 1500 can be implemented in an assembly that includes wearable and communication devices.

Devices 1400 and 1500 can be implemented in an assembly that includes wearable and wireless communication devices.

FIG. 16 shows a flowchart 1600 showing a method of classifying user activities according to an exemplary embodiment. In step 1602, accelerometer data for a plurality of axes is measured. In step 1604, the most active axis of the plurality of axes is identified based on the accelerometer data. In step 1606, user activity is categorized based on the signal amplitude of the accelerometer data of the most active axis and one or more thresholds.

The method can include applying smoothing to accelerometer data before identifying the most active axis.

The method can include identifying the most active axis by finding the axis with the maximum average absolute value of the smoothed signal. This maximum average absolute value is sometimes called the maximum average absolute moving average signal value. The method can include classifying activity based on the maximum signal amplitude of the accelerometer data for the most active axis.

The method can include classifying activity based on the maximum signal amplitude of the accelerometer data for the most active axis minus the average signal amplitude.

The method can include identifying the most active axis based on accelerometer data measured over a predetermined processing window and classifying the user's activity.

The processing window can range from about 0 (not including 0) seconds to 5 seconds.

FIG. 17 shows a flowchart 1700 showing a method of counting the number of steps taken by the user. In step 1702, accelerometer data for at least one axis is measured. In step 1704, the differential operator is applied to the accelerometer data in the continuous processing window. At step 1706, the peak of the derivative of the accelerometer data in each processing window is counted. In step 1708, the overcounted peaks are deleted based on the time difference between the first peak in the current window and the last peak in the preprocessing window.

The method includes counting peaks based on a first threshold when the current processing window is the first processing window after transitioning from a low activity state to a medium activity state or a high activity state. Can be done.

The method can further include updating the adaptive threshold of the next window. The adaptive threshold of the next window can be based on the difference in signal amplitude of the last consecutive peak and valley pair in the current processing window. The adaptive threshold for the next window can be about 0 (not including 0)% to 50% of the difference in signal amplitude between the last consecutive peak and valley pair in the current processing window.

This method removes a peak from the peak count in the current processing window if the time difference between the first peak in the current processing window and the last peak in the preceding processing window is less than the second threshold. Can be included. This second threshold can be about 1/4 second.

Derivative operators can include Pantompukins derivative operators.

An exemplary embodiment of the invention advantageously does not require the device to be mounted in a fixed orientation throughout the monitoring period. This is because the device in the exemplary embodiment constantly detects the most active axis (the axis closest to the direction of impact force) used in subsequent processes within the processing window. This most active axis can be identified by looking for the axis with the maximum average absolute moving average acceleration signal in each window in the exemplary embodiment.

Embodiments of the invention can advantageously make changes detectable, divide the accelerometer data into process windows, and within each window the signal amplitude (eg, 0.1 g and 1 g) relative to the threshold level (eg, 0.1 g and 1 g). For example, by observing the maximum-average or maximum-minimum), it is possible to distinguish the signals corresponding to the low state, the medium state, or the high state.

In embodiments of the present invention, operating noise can advantageously be accurately removed using a low-pass filter with a cutoff frequency suitable for the activity.

In an embodiment of the invention, the Pantonpukins derivative operation involves a peak removal algorithm based on a peak distance and an adaptive threshold based on the maximum-minimum value of the peak, thereby advantageously causing the device and method to peak. Helps identify accurately as a step count.

One of ordinary skill in the art can make numerous modifications and / or modifications to the invention shown in a particular embodiment without departing from the comprehensive description of the invention or scope. Is understood. Therefore, this embodiment is considered to be exemplary and not limiting in all respects. The present invention also provides any combination of features, in particular any combination of features within the scope of the claims, even if the features or combinations of features are not explicitly stated in the claims or embodiments. Including.

For example, wrist-worn devices have been described in some embodiments, which can be worn on the user's arms, hips, waist or feet. In addition, the device and the method can execute only one of the activity classification and the step count as described above, or both can be executed at the same time.

Claims (19)

The device that counts the number of steps taken by the user wearing the device.
It is equipped with an accelerometer for measuring accelerometer data of at least one axis and a processor.
The processor applies a differential operator to the accelerometer data in successive processing windows, counts the peaks of the derivatives of the accelerometer data in each processing window, and with the first peak in the current processing window. A device that is configured to remove overcounted peaks based on the time difference from the last peak in the preprocessing window. The processor is configured to count the peaks based on a first threshold if the current processing window is the first processing window after transitioning from a low activity state to a medium activity state or a high activity state. The device according to claim 1. The device of claim 1 or 2, wherein the processor is further configured to update the adaptive threshold of the next processing window. The device of claim 3, wherein the adaptive threshold of the next processing window is based on the difference in signal amplitude of the last continuous peak and valley pair in the current processing window. The adaptive threshold of the next processing window is about 0 (not including 0)% to 50% of the difference in signal amplitude of the last continuous peak and valley pair in the current processing window. , The device of claim 4. When the time difference between the first peak in the current processing window and the last peak in the preceding processing window is smaller than the second threshold value, the processor causes the peak in the current processing window. The device of any one of claims 1-5, which is configured to remove peaks from the count. The device of claim 6, wherein the second threshold is about 1/4 second. The device according to any one of claims 1 to 7, wherein the differential operator includes a Pantonpukins differential operator. The device according to any one of claims 1 to 8, wherein the device is mounted on a wearable device. The device according to any one of claims 1 to 9, wherein the device is mounted in an assembly including a wearable device and a communication device. The device according to any one of claims 1 to 10, wherein the device is mounted in an assembly including a wearable device and a wireless communication device. It is a method of counting the number of steps taken by the user.
Measuring accelerometer data for at least one axis and
Applying a differential operator to the accelerometer data in a continuous processing window,
Counting the peak of the derivative of the accelerometer data in each processing window,
Removing overcounted peaks based on the time difference between the first peak in the current processing window and the last peak in the preprocessing window,
Including methods. 12. If the current processing window is the first processing window after transitioning from a low activity state to a medium activity state or a high activity state, claim 12 includes counting the peak based on a first threshold. The method described. 12. The method of claim 12 or 13, further comprising updating the adaptive threshold of the next processing window. 14. The method of claim 14, wherein the adaptive threshold of the next processing window is based on the difference in signal amplitude of the last continuous peak and valley pair in the current processing window. The adaptive threshold of the next processing window is about 0 (not including 0)% to 50% of the difference in signal amplitude of the last continuous peak and valley pair in the current processing window. , The method of claim 15. When the time difference between the first peak in the current processing window and the last peak in the preceding processing window is smaller than the second threshold, the peak is calculated from the peak count in the current processing window. The method according to any one of claims 12 to 16, comprising deleting. The method of claim 17, wherein the second threshold is about 1/4 second. The method according to any one of claims 12 to 18, wherein the differential operator includes a Pantonpukins differential operator.

Images