A motion state recognition method and device based on wearable device

Abstract

The present invention discloses a motion state recognition method and device based on a wearable device, the method comprising: obtaining three-axis acceleration data in a short time domain, and performing low-pass filtering on the three-axis acceleration data; determining the acceleration axis in the direction of gravity according to the three-axis acceleration data after low-pass filtering, and judging whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the direction of gravity; when the current state is not the first state, calculating the combined acceleration in the short time domain according to the three-axis acceleration data, and judging whether the current motion state is the second state according to the extreme difference of the combined acceleration in the short time domain; when the current state is not the second state, calculating the spectrum data of the combined acceleration, and judging the current motion state according to the power proportion of the spectrum data in a preset power frequency band. The present invention solves the technical problems of large system load and high implementation difficulty in current motion state recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent wearable devices, and in particular to a motion state recognition method and device based on a wearable device, a wearable device and a storage medium. Background Art

[0002] Wearable devices can support a variety of sports, including walking, race walking, running, cycling, swimming, etc. Most products on the market require users to choose their own sports methods. In the free movement state, it is impossible to solve the movement state in time. Nowadays, machine learning methods are commonly used algorithms for motion state recognition, but the model construction of such algorithms (offline work) requires a lot of training time, and the feedforward deployment of deep learning is computationally intensive, resulting in a large system load and high implementation difficulty. Summary of the invention

[0003] The purpose of the present invention is to overcome the above-mentioned technical deficiencies and provide a motion state recognition method, apparatus, wearable device and storage medium based on a wearable device, so as to solve the technical problems of large system load and high implementation difficulty caused by using deep learning method for motion state recognition in the prior art.

[0004] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a method for identifying a motion state based on a wearable device, comprising the following steps:

[0006] Acquire three-axis acceleration data in a short time domain, and perform low-pass filtering on the three-axis acceleration data;

[0007] Determine the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and judge whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction;

[0008] When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data, and according to the range of the combined acceleration in the short time domain, it is determined whether the current motion state is the second state;

[0009] When the current state is not the second state, the spectrum data of the combined acceleration is calculated, and the current motion state is determined according to the power proportion of the spectrum data in a preset power frequency band.

[0010] Preferably, in the motion state identification method based on a wearable device, the motion state includes at least a swimming state, a cycling state, a walking state, a race walking state and a running state, the first state is a swimming state, and the second state is a cycling state.

[0011] Preferably, in the motion state recognition method based on the wearable device, the acceleration axis in the gravity direction is determined according to the three-axis acceleration data after low-pass filtering, and the current motion state is judged as the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction, including:

[0012] Perform empirical mode decomposition on the three-axis acceleration data after low-pass filtering to obtain the baseline value of the three-axis acceleration;

[0013] According to the baseline values ​​of the three-axis acceleration, the acceleration axis in the direction of gravity is determined;

[0014] Perform fast Fourier transform on the three-axis acceleration data after low-pass filtering to obtain the Fourier series of the three-axis acceleration;

[0015] The power proportion of the acceleration of the acceleration axis in the direction of gravity is calculated based on the Fourier series of the three-axis acceleration;

[0016] Based on the power proportion of the acceleration of the acceleration axis in the gravity direction, it is determined whether the current state is the first state.

[0017] Preferably, in the motion state recognition method based on the wearable device, judging whether the current state is the first state based on the power proportion of the acceleration of the acceleration axis in the gravity direction includes:

[0018] It is determined whether the power proportion of the acceleration of the acceleration axis on the gravity exceeds a first preset value. If so, it is determined that the current state is not the first state; otherwise, it is determined that the current state is the first state.

[0019] Preferably, in the motion state identification method based on the wearable device, when the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data, and according to the range of the combined acceleration in the short time domain, whether the current motion state is the second state is determined, including:

[0020] When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data;

[0021] Calculate the maximum and minimum values ​​of the combined acceleration in the short time domain, and calculate the extreme difference value of the combined acceleration in the short time domain;

[0022] When the range value is less than the second preset value, the current state is determined to be the second state; otherwise, the current state is determined not to be the second state.

[0023] Preferably, in the motion state identification method based on the wearable device, when the current state is not the second state, the spectrum data of the combined acceleration is calculated, and the current motion state is determined according to the power proportion of the spectrum data in a preset power frequency band, including:

[0024] When the current state is not the second state, performing a fast Fourier transform on the combined acceleration to obtain frequency spectrum data of the combined acceleration in a short time domain;

[0025] The current motion state is determined according to the power ratio of the spectrum data in a preset power frequency band.

[0026] Preferably, in the motion state identification method based on a wearable device, judging the current motion state according to the power proportion of the spectrum data in a preset power frequency band includes:

[0027] When the power proportion of the spectrum data in the first power frequency band is the highest, determining that the current motion state is the third state;

[0028] When the power proportion of the spectrum data in the second power frequency band is the highest, determining that the current motion state is the fourth state;

[0029] When the power proportion of the spectrum data in the third power frequency band is the highest, it is determined that the current motion state is the fifth state.

[0030] In a second aspect, the present invention further provides a motion state recognition device based on a wearable device, comprising:

[0031] A filtering module, used to obtain three-axis acceleration data in a short time domain and perform low-pass filtering on the three-axis acceleration data;

[0032] A first judgment module is used to determine the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and to judge whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction;

[0033] The second judgment module is used to calculate the combined acceleration in the short time domain according to the three-axis acceleration data when the current state is not the first state, and judge whether the current motion state is the second state according to the range of the combined acceleration in the short time domain;

[0034] The third judgment module is used to calculate the spectrum data of the combined acceleration when the current state is not the second state, and judge the current motion state according to the power proportion of the spectrum data in a preset power frequency band.

[0035] In a third aspect, the present invention further provides a wearable device, comprising an accelerometer and an electronic device, wherein the electronic device comprises a processor and a memory, wherein the accelerometer is electrically connected to the processor and is used to send the collected triaxial acceleration data in a short time domain to the processor;

[0036] The memory stores a computer-readable program executable by the processor;

[0037] When the processor executes the computer-readable program, the steps in the motion state recognition method based on the wearable device as described above are implemented.

[0038] In a third aspect, the present invention also provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the motion state recognition method based on a wearable device as described above.

[0039] Compared with the prior art, the motion state recognition method, device, wearable device and storage medium based on wearable device provided by the present invention abandon the framework of machine learning and innovatively use the power ratio of each frequency band as the standard for judging the motion state. First, the user collects acceleration data through the chest-mounted device, and smoothes and filters the acceleration to obtain an acceleration time queue with a high signal-to-noise ratio; then, the acceleration power ratio of the three axes of acceleration is solved to determine whether it is the first state; if it is not the first state, the acceleration range is first calculated to determine whether it is the second state, and then the acceleration sequence after filtering and smoothing is fast Fourier transformed to obtain the corresponding spectrum data; then, the power ratio of each frequency band is obtained according to the spectrum data; finally, the user's motion state is identified by the power ratio of each frequency band, and the motion state recognition using the non-machine learning framework is realized, which saves the process of offline training parameters and has good real-time performance. The time series processed is at most two motion state cycles, the calculation delay is short, and other equipment does not need to be connected, only a wearable device with an accelerometer is needed, which saves costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of a preferred embodiment of the motion state recognition method based on a wearable device provided by the present invention;

[0041] Figure 2a It is a box plot of the power proportion of the gravity axis in various motion states;

[0042] Figure 2b It is a time series diagram of the power proportion of the gravity axis in various motion states;

[0043] Figure 3 It is a schematic diagram of the acceleration amplitude in the non-swimming state;

[0044] Figure 4a This is a time domain analysis diagram in a fast running state;

[0045] Figure 4b This is a time domain analysis diagram under the race walking state;

[0046] Figure 4c It is a time domain analysis diagram under walking state;

[0047] Figure 5 It is the frequency domain analysis diagram under the states of fast running, race walking and slow walking;

[0048] Figure 6 is a schematic diagram of an embodiment of a motion state recognition device based on a wearable device provided by the present invention;

[0049] Figure 7 It is a schematic diagram of the operating environment of an embodiment of the motion state recognition program based on a wearable device of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0051] See also Figure 1 The embodiment of the present invention provides a method for identifying a motion state based on a wearable device, comprising the following steps:

[0052] S100, acquiring three-axis acceleration data in a short time domain, and performing low-pass filtering on the three-axis acceleration data;

[0053] S200, determining the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and judging whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction;

[0054] S300, when the current state is not the first state, calculating the combined acceleration in the short time domain according to the three-axis acceleration data, and judging whether the current motion state is the second state according to the range of the combined acceleration in the short time domain;

[0055] S400: When the current state is not the second state, the spectrum data of the combined acceleration is calculated, and the current motion state is determined according to the power proportion of the spectrum data in a preset power frequency band.

[0056] In this embodiment, the framework of machine learning is abandoned, and the power ratio of each frequency band is innovatively used as a standard for judging the state of motion. First, the user collects acceleration data through a chest-mounted device, and smoothes and filters the acceleration to obtain an acceleration time queue with a high signal-to-noise ratio; then, the acceleration power ratio of the three axes of acceleration is solved to determine whether it is the first state; if it is not the first state, the acceleration range is first calculated to determine whether it is the second state, and then the acceleration sequence after filtering and smoothing is fast Fourier transformed to obtain the corresponding spectrum data; then, the power ratio of each frequency band is obtained according to the spectrum data; finally, the user's motion state is identified through the power ratio of each frequency band, and the motion state recognition using a non-machine learning framework is realized, which saves the process of offline training parameters and has good real-time performance. The time series processed is at most two motion state cycles, the calculation delay is short, and no other equipment is required to access, only a wearable device with an accelerometer is required, which saves costs.

[0057] In some embodiments, the motion state includes at least a swimming state, a cycling state, a walking state, a race walking state, and a running state, wherein the first state is a swimming state, and the second state is a cycling state. The motion state can be accurately identified by the above motion state recognition method. It should be noted that the swimming state, cycling state, walking state, race walking state, and running state described in the embodiments of the present invention are all user behavior states, which does not mean that the user must be swimming, cycling, etc. For example, the user can simulate a running state through a treadmill, or make swimming-like movements, but is not actually swimming.

[0058] In some embodiments, step S100 is to improve the signal-to-noise ratio of the acceleration signal, thereby obtaining an acceleration signal with a higher signal-to-noise ratio. In specific implementation, the obtained acceleration sequence (three-axis data) is first subjected to 5 Hz low-pass sampling processing to improve the signal-to-noise ratio of the acceleration signal:

[0059]

[0060] Among them, y(n) is the nth acceleration value of the filter output, and x is the input acceleration sequence. In order to distinguish the three axes, y is calculated. i (i=1,2,3) represents three acceleration axes.

[0061] In some embodiments, step S200 is to determine whether the current motion state is a swimming state. Specifically, step S200 specifically includes:

[0062] Perform empirical mode decomposition on the three-axis acceleration data after low-pass filtering to obtain the baseline value of the three-axis acceleration;

[0063] According to the baseline values ​​of the three-axis acceleration, the acceleration axis in the direction of gravity is determined;

[0064] Perform fast Fourier transform on the three-axis acceleration data after low-pass filtering to obtain the Fourier series of the three-axis acceleration;

[0065] The power proportion of the acceleration of the acceleration axis in the direction of gravity is calculated based on the Fourier series of the three-axis acceleration;

[0066] Based on the power proportion of the acceleration of the acceleration axis in the gravity direction, it is determined whether the current state is the first state.

[0067] In this embodiment, the acceleration yi of the three axes is first subjected to empirical mode decomposition. Empirical mode decomposition is a new adaptive signal time-frequency processing method creatively proposed by NE Huang and others at NASA in 1998, which is particularly suitable for the analysis and processing of nonlinear and non-stationary signals. The baseline value of the three-axis acceleration can be obtained through empirical mode decomposition, and the baseline value is closest to g (gravitational acceleration). It is considered that the direction of the axis is roughly in the same direction or opposite direction as gravity, so the acceleration axis in the direction of gravity can be obtained. After the acceleration axis of your in the direction of gravity is determined, the acceleration of the three axes can be subjected to fast Fourier transform, and the following is obtained:

[0068]

[0069] Where F represents the Fourier series sequence of acceleration, i is a unit complex number, i 2 = -1, n represents the labels of the three axes. Based on this series, the power ratio of the three axes is calculated:

[0070] fea=fea1+fea2+fea3

[0071]

[0072]

[0073]

[0074] Since in the previous step, the axis that is approximately collinear with the gravitational acceleration has been determined, and the acceleration ratio of the axis has been obtained, denoted as P per , through P per The size of the icon can quickly determine whether the current motion state is swimming.

[0075] In some embodiments, judging whether the current state is the first state based on the power proportion of the acceleration of the acceleration axis in the gravity direction includes:

[0076] It is determined whether the power proportion of the acceleration of the acceleration axis on the gravity exceeds a first preset value. If so, it is determined that the current state is not the first state; otherwise, it is determined that the current state is the first state.

[0077] See also Figure 2a and Figure 2b In this embodiment, an empirical value is obtained through analysis of a large amount of motion data, namely the first preset value mentioned above, which is preferably 50%. When the power proportion threshold of the gravity axis is above 50%, the current motion state can be determined as a non-swimming state, otherwise the current motion state can be determined as a swimming state.

[0078] In some embodiments, step S300 is to determine whether the current state is a cycling state when the current motion state is not a swimming state. Specifically, step S300 specifically includes:

[0079] When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data;

[0080] Calculate the maximum and minimum values ​​of the combined acceleration in the short time domain, and calculate the extreme difference value of the combined acceleration in the short time domain;

[0081] When the range value is less than the second preset value, the current state is determined to be the second state; otherwise, the current state is determined not to be the second state.

[0082] In this embodiment, if the exercise state is determined to be a non-swimming state, the system enters the next level of judgment, first judging the extreme difference of acceleration. If it does not reach the judgment threshold (i.e., the second preset value), it is a cycling state. If it exceeds the threshold, it is a walking, race walking or running state. Please refer to Figure 3 ,from Figure 3 It can be seen that the extreme difference in acceleration during cycling is around 0.5g, while there is overlap in acceleration amplitudes for running, race walking, and slow walking. Therefore, if the acceleration amplitude is less than 0.5g, it is judged as cycling, otherwise it is walking, race walking, or running.

[0083] In some embodiments, step S400 is to determine whether the current motion state is a walking state, a race walking state or a running state when the current motion state is neither a swimming state nor a cycling state. Specifically, step S400 specifically includes:

[0084] When the current state is not the second state, performing a fast Fourier transform on the combined acceleration to obtain frequency spectrum data of the combined acceleration in a short time domain;

[0085] The current motion state is determined according to the power ratio of the spectrum data in a preset power frequency band.

[0086] In this embodiment, as with the above-mentioned Fourier transform, when the current state is not the riding state, the acceleration is subjected to a fast Fourier transform to obtain the spectrum data of the combined acceleration in the short time domain, and the power of 0.5-1.5Hz, 1.5-2.5Hz, 2.5-3.5Hz, and 3.5-5Hz is calculated, such as Figure 4a , Figure 4b , Figure 4c and Figure 5 As shown, the short-time domain spectrum analysis is as follows:

[0087] ①

[0088] ②

[0089] ③

[0090] ④

[0091] The above formula can be used to calculate the proportion of the power of 0.5-1.5Hz, 1.5-2.5Hz, 2.5-3.5Hz, and 3.5-5Hz in the power of 0.5-5Hz. Next, the correlation coefficient between the ratio vector and the diagonal unit matrix is ​​solved to determine the motion state.

[0092] In some embodiments, judging the current motion state according to the power proportion of the spectrum data in a preset power frequency band includes:

[0093] When the power proportion of the spectrum data in the first power frequency band is the highest, determining that the current motion state is the third state;

[0094] When the power proportion of the spectrum data in the second power frequency band is the highest, determining that the current motion state is the fourth state;

[0095] When the power proportion of the spectrum data in the third power frequency band is the highest, it is determined that the current motion state is the fifth state.

[0096] In this embodiment, if the power proportion of the ② frequency band is the highest, it is a walking state; if the power proportion of the ③ frequency band is the highest, it is a race walking state; if the power proportion of the ④ frequency band is the highest, the exercise state is a running state.

[0097] The present invention uses a non-machine learning framework to perform motion state recognition, saving the process of offline training parameters. Compared with traditional machine learning algorithms, the accuracy of motion state recognition of the present invention is not much different, but the system is lightweight and can greatly reduce the difficulty of implementing the system.

[0098] Based on the above-mentioned motion state recognition method based on a wearable device, the embodiment of the present invention also provides a motion state recognition device 600 based on a wearable device. Figure 6 The motion state recognition device based on the wearable device includes a filtering module 610, a first judgment module 620, a second judgment module 630 and a third judgment module 640.

[0099] The filtering module 610 is used to obtain the three-axis acceleration data in the short time domain and perform low-pass filtering on the three-axis acceleration data;

[0100] The first judgment module 620 is used to determine the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and determine whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction;

[0101] The second judgment module 630 is used to calculate the combined acceleration in the short-time domain according to the three-axis acceleration data when the current state is not the first state, and judge whether the current motion state is the second state according to the range of the combined acceleration in the short-time domain;

[0102] The third judgment module 640 is used to calculate the spectrum data of the combined acceleration when the current state is not the second state, and judge the current motion state according to the power proportion of the spectrum data in a preset power frequency band.

[0103] In this embodiment, the framework of machine learning is abandoned, and the power ratio of each frequency band is innovatively used as a standard for judging the state of motion. First, the user collects acceleration data through a chest-mounted device, and smoothes and filters the acceleration to obtain an acceleration time queue with a high signal-to-noise ratio; then, the acceleration power ratio of the three axes of acceleration is solved to determine whether it is the first state; if it is not the first state, the acceleration range is first calculated to determine whether it is the second state, and then the acceleration sequence after filtering and smoothing is fast Fourier transformed to obtain the corresponding spectrum data; then, the power ratio of each frequency band is obtained according to the spectrum data; finally, the user's motion state is identified through the power ratio of each frequency band, and the motion state recognition using a non-machine learning framework is realized, which saves the process of offline training parameters and has good real-time performance. The time series processed is at most two motion state cycles, the calculation delay is short, and no other equipment is required to access, only a wearable device with an accelerometer is required, which saves costs.

[0104] Since the motion state recognition method based on a wearable device has been described in detail above, the technical effects of the motion state recognition method based on a wearable device are also possessed by the motion state recognition device based on a wearable device, so they will not be repeated here.

[0105] like Figure 7 As shown, based on the above-mentioned motion state recognition method based on wearable device, the present invention also provides a wearable device, including an accelerometer and an electronic device, the electronic device including a processor and a memory, the accelerometer is electrically connected to the processor and is used to send the collected three-axis acceleration data in the short time domain to the processor.

[0106] The electronic device may be a computing device such as a controller, a mobile terminal, a desktop computer, a notebook, a palm computer, a server, etc. The electronic device includes a processor 10 , a memory 20 and a display 30 . Figure 7 Only some components of the electronic device are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0107] In some embodiments, the memory 20 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. In other embodiments, the memory 20 may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 20 may also include both an internal storage unit of the electronic device and an external storage device. The memory 20 is used to store application software and various types of data installed in the electronic device, such as program codes installed in the electronic device. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a motion state recognition program 40 based on a wearable device is stored on the memory 20, and the motion state recognition program 40 based on the wearable device can be executed by the processor 10, thereby realizing the motion state recognition method based on the wearable device of each embodiment of the present invention.

[0108] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 20, such as executing a motion state recognition method based on a wearable device.

[0109] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the electronic device and to display a visual user interface. The components 10-30 of the electronic device communicate with each other via a system bus.

[0110] In some embodiments, when the processor 10 executes the motion state recognition program 40 based on the wearable device in the memory 20, the steps of the motion state recognition method based on the wearable device described in the above embodiments are implemented.

[0111] In summary, the motion state recognition method, device, wearable device and storage medium based on wearable device provided by the present invention abandon the framework of machine learning and innovatively use the power ratio of each frequency band as the standard for judging the motion state. First, the user collects acceleration data through the chest-mounted device, and smoothes and filters the acceleration to obtain an acceleration time queue with a high signal-to-noise ratio; then, the acceleration power ratio of the three axes of acceleration is solved to determine whether it is the first state; if it is not the first state, the acceleration range is first calculated to determine whether it is the second state, and then the acceleration sequence after filtering and smoothing is fast Fourier transformed to obtain the corresponding spectrum data; then, the power ratio of each frequency band is obtained according to the spectrum data; finally, the user's motion state is identified by the power ratio of each frequency band, and the motion state recognition using the non-machine learning framework is realized, which saves the process of offline training parameters and has good real-time performance. The time series processed is at most two motion state cycles, the calculation delay is short, and other equipment does not need to be connected, only a wearable device with an accelerometer is needed, which saves costs.

[0112] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The storage medium can be a memory, a disk, an optical disk, etc.

[0113] The specific implementation of the present invention described above does not constitute a limitation on the protection scope of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A motion state recognition method based on a wearable device, characterized in that: The steps include: Acquire three-axis acceleration data in a short time domain, and perform low-pass filtering on the three-axis acceleration data; Determine the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and judge whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction; When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data, and according to the range of the combined acceleration in the short time domain, it is determined whether the current motion state is the second state; When the current state is not the second state, the spectrum data of the combined acceleration is calculated, and the current motion state is determined according to the power proportion of the spectrum data in a preset power frequency band.

2. The motion state recognition method based on a wearable device according to claim 1, characterized in that: The motion state includes at least a swimming state, a cycling state, a walking state, a race walking state and a running state. The first state is a swimming state, and the second state is a cycling state.

3. The motion state recognition method based on a wearable device according to claim 1, characterized in that: The step of determining the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and judging whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction, includes: Perform empirical mode decomposition on the three-axis acceleration data after low-pass filtering to obtain the baseline value of the three-axis acceleration; According to the baseline values ​​of the three-axis acceleration, the acceleration axis in the direction of gravity is determined; Perform fast Fourier transform on the three-axis acceleration data after low-pass filtering to obtain the Fourier series of the three-axis acceleration; The power proportion of the acceleration of the acceleration axis in the direction of gravity is calculated based on the Fourier series of the three-axis acceleration; Based on the power proportion of the acceleration of the acceleration axis in the gravity direction, it is determined whether the current state is the first state.

4. The motion state recognition method based on a wearable device according to claim 3, characterized in that: The determining whether the current state is the first state based on the power proportion of the acceleration of the acceleration axis in the gravity direction includes: It is determined whether the power proportion of the acceleration of the acceleration axis in the gravity direction exceeds a first preset value. If so, it is determined that the current state is not the first state; otherwise, it is determined that the current state is the first state.

5. The motion state recognition method based on a wearable device according to claim 1, characterized in that: When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data, and according to the range of the combined acceleration in the short time domain, whether the current motion state is the second state is determined, including: When the current state is not the first state, the combined acceleration in the short time domain is calculated according to the three-axis acceleration data; Calculate the maximum and minimum values ​​of the combined acceleration in the short time domain, and calculate the extreme difference value of the combined acceleration in the short time domain; When the range value is less than the second preset value, the current state is determined to be the second state; otherwise, the current state is determined not to be the second state.

6. The motion state recognition method based on a wearable device according to claim 1, characterized in that: When the current state is not the second state, the spectrum data of the combined acceleration is calculated, and the current motion state is determined according to the power proportion of the spectrum data in a preset power frequency band, including: When the current state is not the second state, performing a fast Fourier transform on the combined acceleration to obtain frequency spectrum data of the combined acceleration in a short time domain; The current motion state is determined according to the power ratio of the spectrum data in a preset power frequency band.

7. The motion state recognition method based on a wearable device according to claim 6, characterized in that: The determining the current motion state according to the power proportion of the spectrum data in a preset power frequency band includes: When the power proportion of the spectrum data in the first power frequency band is the highest, determining that the current motion state is the third state; When the power proportion of the spectrum data in the second power frequency band is the highest, determining that the current motion state is the fourth state; When the power proportion of the spectrum data in the third power frequency band is the highest, it is determined that the current motion state is the fifth state.

8. A motion state recognition device based on a wearable device, characterized in that: include: A filtering module, used to obtain three-axis acceleration data in a short time domain and perform low-pass filtering on the three-axis acceleration data; A first judgment module is used to determine the acceleration axis in the gravity direction according to the three-axis acceleration data after low-pass filtering, and to judge whether the current motion state is the first state according to the power proportion of the acceleration of the acceleration axis in the gravity direction; The second judgment module is used to calculate the combined acceleration in the short time domain according to the three-axis acceleration data when the current state is not the first state, and judge whether the current motion state is the second state according to the range of the combined acceleration in the short time domain; The third judgment module is used to calculate the spectrum data of the combined acceleration when the current state is not the second state, and judge the current motion state according to the power proportion of the spectrum data in a preset power frequency band.

9. A wearable device, characterized in that: The device comprises an accelerometer and an electronic device, wherein the electronic device comprises a processor and a memory, and the accelerometer is electrically connected to the processor and is used to send the collected triaxial acceleration data in a short time domain to the processor; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the motion state recognition method based on a wearable device as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the motion state recognition method based on a wearable device as described in any one of claims 1-7.

Images