TW202025107A - Sports teaching assisted system based on wearable device - Google Patents

Abstract

A sport teaching assisted system based on a wearable device includes the wearable device and a processing unit. The wearable device is worn on an arm of a user. The wearable device includes a gyroscope and at least one electromyography sensor respectively used to sense a 3-axis angular velocity and at least one electromyography signal corresponding to different time points when the user performs an arm swing motion. The processing unit receives the 3-axis angular velocities and the electromyography signals so as to calculate a swing gesture eigenvalue and a muscle strength eigenvalue. The processing unit input the swing gesture eigenvalue and the muscle strength eigenvalue into a standard sports model, thereby calculating a percentage of swing gesture similarity and a percentage of muscle strength similarity.

Description

Sports Teaching Assistant Department Based on Wearable Devices Unify

The disclosed embodiment relates to an exercise teaching assist system, and particularly relates to a wearable device-based exercise teaching assist system.

In the process of sports training, the coach usually uses one-to-one or one-to-many teaching mode. Because the teaching is conducted manually, a coach must be watched from the side to determine whether the athlete’s posture or action Meet the requirements, time-consuming and laborious. Furthermore, when athletes perform personal exercises after teaching, they cannot judge whether their postures and movements meet the requirements, which greatly reduces the efficiency of personal exercises. Therefore, it is necessary to develop a sports teaching aid system to solve the above problems.

The purpose of this disclosure is to provide a wearable device-based exercise teaching assistance system, which includes a wearable device and a processing unit. The wearable device is used for the user to wear on the arm. The wearable device includes a three-axis spinning top The meter and at least one EMG (Electromyography) sensor are used to sense the three-axis angular velocity values and at least one EMG signal corresponding to different time points when the user makes an arm swing. The processing unit is used to receive the three-axis angular velocity value and the electromyographic signal to calculate the posture feature value and the force feature value corresponding to the arm swing, and input the posture feature value and the force feature value into the standard motion model to calculate the posture similarity The percentage of degree and the percentage of similarity of force.

In some embodiments, the aforementioned processing unit uses Riemann Integral to convert the three-axis angular velocity values into posture feature values and normalize them, and then input the normalized posture feature values into the standard motion model to Calculate the percentage of posture similarity.

In some embodiments, the above-mentioned processing unit converts the myoelectric signal into a force characteristic value by means of a mean absolute value (Mean Absolute Value) operation and normalizes the force characteristic value, and then inputs the normalized force characteristic value into the standard motion model. To calculate the force similarity percentage.

In some embodiments, before the processing unit inputs the posture feature value and the force feature value into the standard motion model, the processing unit divides the arm swing action into multiple decompositions according to changes in the three-axis angular velocity value with different time points Action, the processing unit is used to calculate the posture feature value and the force feature value corresponding to each decomposition action.

In some embodiments, the process of establishing the above-mentioned standard motion model is as follows: wear a wearable device on the arm of each coach; the wearable device senses that when each coach makes multiple standard arm swings, each standard swing is The actions correspond to the three-axis angular velocity values at different time points and at least one EMG signal; the processing unit divides each standard arm swing into multiple standard decomposition movements The processing unit uses Riemann integration to convert the three-axis angular velocity value corresponding to each standard decomposition action into a posture feature value and normalizes it; the processing unit uses average absolute value operation to correspond to each standard decomposition action The EMG signal is converted into force eigenvalues and normalized; and the processing unit uses the normalized posture eigenvalues and force eigenvalues corresponding to each standard decomposition action as Back Propagation Artificial Neural Network, BPANN) training data to train the standard motion model.

In some embodiments, the above-mentioned processing unit divides each standard arm swing action into a standard decomposition action according to changes of the three-axis angular velocity value with different time points.

In some embodiments, after the standard motion model is established, the processing unit further determines the classification accuracy rate of the standard motion model by 10-fold cross-validation.

In some embodiments, the wearable device-based exercise teaching assistance system further includes a man-machine interface for presenting the posture similarity percentage and the force similarity percentage corresponding to each decomposition action through the radar chart interface.

In some embodiments, the above-mentioned human-machine interface further includes a teaching suggestion interface for providing action correction suggestions to the user according to the posture similarity percentage and the force similarity percentage corresponding to each decomposition action.

In some embodiments, the above-mentioned human-machine interface further includes a coach video menu and a user video menu. The user clicks the coach video menu and the user video menu to play videos for action comparison.

In order to make the above-mentioned features and advantages of the present disclosure more obvious and understandable, the following specific embodiments are described in detail in conjunction with the accompanying drawings.

100‧‧‧A sports teaching aid system based on wearable devices

120‧‧‧Wearable device

122‧‧‧Three-axis gyroscope

124‧‧‧Myoelectricity sensor

140‧‧‧Processing unit

160‧‧‧Human Machine Interface

162‧‧‧Posture similarity radar chart

164‧‧‧Strength similarity radar chart

166‧‧‧teaching suggestion interface

168‧‧‧Video Menu

P1~P9‧‧‧point

From the following detailed description in conjunction with the accompanying drawings, a better understanding of the aspect of the disclosure can be obtained. It should be noted that, according to industry standard practice, each feature is not drawn to scale. In fact, in order to make the discussion clearer, the size of each feature can be arbitrarily increased or decreased.

[Fig. 1] is a schematic diagram of an exercise teaching assist system based on a wearable device according to an embodiment of the disclosure.

[Figure 2] is a schematic diagram of a wearable device according to an embodiment of the disclosure.

[Fig. 3] is a schematic diagram of the arm swing action divided into four decomposition actions according to the embodiment of the disclosure.

[Fig. 4] is a schematic diagram of the display screen of the human-machine interface according to the embodiment of the disclosure.

The embodiments of the present invention are discussed in detail below. However, it can be understood that the embodiments provide many applicable concepts, which can be implemented in various specific contents. The discussed and disclosed embodiments are for illustration only, and are not intended to limit the scope of the present invention.

1 is a schematic diagram of a wearable device-based exercise teaching assistance system 100 according to an embodiment of the present disclosure. The wearable device-based exercise teaching assistance system 100 includes a wearable device 120 and a processing unit 140. The wearable device 120 is used for the user to wear on the arm to detect the user's arm swing.

FIG. 2 is a schematic diagram of a wearable device 120 according to an embodiment of the present disclosure. The wearable device 120 includes a three-axis gyroscope (Gyroscopes) 122 and an electromyography (EMG) sensor 124. The three-axis gyroscope 122 is used to sense the action angle when the user makes an arm swing, which means that the three-axis gyroscope 122 is used to sense the three-axis angular velocity values corresponding to different time points when the user makes an arm swing. Among them, the three-axis angular velocity value includes the X-axis angular velocity value, the Y-axis angular velocity value and the Z-axis angular velocity value. The myoelectric sensor 124 is used to sense the muscle strength when the user makes an arm swing. That is, the myoelectric sensor 124 is used to sense the myoelectric signals corresponding to different time points when the user makes an arm swing. In the disclosed embodiment, the wearable device 120 includes eight myoelectric sensors 124 covering eight different parts of the user's arm to sense the eight different time points when the user makes an arm swing. The myoelectric signal can more completely sense the muscle strength when the user makes an arm swing. However, the present disclosure is not limited to this. The designer can choose the number of myoelectric sensors 124 included in the wearable device 120 according to actual needs. .

Please return to FIG. 1, the processing unit 140 receives from the wearable device 120 the three-axis angular velocity values and myoelectric signals corresponding to different time points when the user makes an arm swing movement through wireless transmission, such as Bluetooth or wireless network. number. Since an arm swing may include different arm swings, for example, the forward shot and kill action of badminton includes four sub-actions in sequence: relaxation before the shot, preparatory actions before the shot, the index finger drives the wrist to buckle, Relax after the shot. If you can break an arm swing into multiple decompositions, it will help you Effectively assist in sports teaching. In the disclosed embodiment, the processing unit 140 divides an arm swing into multiple decomposition actions according to the changes of the three-axis angular velocity values at different time points. Once the division is completed, the processing unit 140 can obtain the results of each decomposition action. Corresponding to the three-axis angular velocity value and EMG signal respectively. The following uses FIG. 3 as an example to illustrate how the present disclosure divides an arm swing action into multiple decomposition actions according to the changes of the three-axis angular velocity values with different time points.

FIG. 3 is a schematic diagram of the arm swing action divided into four decomposition actions according to the embodiment of the disclosure. Figure 3 shows a schematic diagram of the Y-axis angular velocity change over time when the arm swing is a forward shot of a shuttlecock. Among the three-axis angular velocity values, the decomposition action 1 is to relax before the shot, and the decomposition action 2 is to kill. Preparatory action before the ball, decomposition action 3 is the index finger drives the wrist to buckle down, and decomposition action 4 is relaxation after the shot. When performing decomposition action 3: When the index finger drives the wrist to buckle down, the user's body quickly turns back to face the net and the arm must smash from top to bottom quickly. Therefore, the Y-axis angular velocity value will suddenly decrease to a negative value, and the minimum Negative values will appear obvious troughs in Figure 3, that is, point P6 in the decomposition action 3. Then find out the starting point and ending point of decomposition action 3. Regarding the starting point of decomposition action 3, first find the "previous maximum value" (ie point P5) from point P6, and then go forward to find the Y axis angular velocity value Point P4 approaching zero, the Y-axis angular velocity value approaching zero means that there is no invention change in the angle. Therefore, it can be judged that point P4 is the starting point of decomposition action 3; regarding the end point of decomposition action 3, first from point P6 to the back Find the "post-maximum value" (ie point P7), and then find the point P8 where the Y-axis angular velocity value changes flat and close to zero. The Y-axis angular velocity value changes flat and close to zero, which means that the angle has not changed. , So it can be judged Breakpoint P8 is the end point of decomposition action 3. When performing decomposition action 1: When relaxing before the shot, the arm must be raised flat on the waist and kept still, so the Y-axis angular velocity value will continue to be kept at a relatively low value; when performing decomposition action 2: preparatory action before the shot , The user must face the net sideways, while raising his arms almost perpendicular to the ground to prepare to hit the ball. According to the above, if you want to find the starting point and ending point of decomposition action 2, the end point of decomposition action 2 is the starting point of decomposition action 3 (ie point P4), and for the starting point of decomposition action 2, first Find the "previous maximum value" from point P4 (ie point P3), and then go forward to find the point P2 where the Y-axis angular velocity value changes flat. The Y-axis angular velocity value changes flat, which means that the arm remains stationary, so it can be judged Point P2 is the starting point of decomposition action 2. If you want to find the starting point and ending point of decomposition action 1, the end point of decomposition action 1 is the starting point of decomposition action 2 (ie point P2), and the starting point of decomposition action 1 is the first point in Figure 3. (Ie point P1). If you want to find the start point and end point of decomposition action 4, the start point of decomposition action 4 is the end point of decomposition action 3 (ie point P8), and the end point of decomposition action 4 is the end point ( Click P9). According to the above judgment process, it is established that the time zone paragraph of decomposition action 1 is between points P1 and P2, the time zone paragraph of decomposition action 2 is between points P2 and P4, and the time zone paragraph of decomposition action 3 is between points P4 and P4. Between point P8, the time zone segment of decomposition action 4 is between point P8 and point P9.

Please return to FIG. 1, the processing unit 140 uses Riemann Integral to convert multiple three-axis angular velocity values corresponding to different time points when the user makes an arm swing motion into posture characteristic values. In the disclosed embodiment, if the processing unit 140 first divides an arm swing action into a plurality of decomposition actions, the processing unit 140 uses Riemann integration to divide each decomposition action to The multiple three-axis angular velocity values at different time points are converted into posture feature values, which means that the four decomposition actions will correspond to the four posture feature values respectively. On the other hand, the processing unit 140 uses Mean Absolute Value (Mean Absolute Value) calculation to convert the multiple EMG signals corresponding to different time points when the user makes an arm swing motion into force characteristic values. In the disclosed embodiment, if the processing unit 140 first divides an arm swing action into multiple decomposition actions, the processing unit 140 uses the average absolute value operation to correspond each decomposition action to multiple myographs at different time points. The number is converted into a force characteristic value, which means that the four decomposition actions will correspond to the four force characteristic values. In the disclosed embodiment, if the wearable device 120 includes a plurality of myoelectric sensors 124, the processing unit 140 uses average absolute value calculation to associate each myoelectric sensor 124 with multiple myoelectric signals at different time points. The signal is converted into a force characteristic value, which means that the eight myoelectric sensors 124 will correspond to the eight force characteristic values respectively.

The processing unit 140 then normalizes the posture feature value so that the posture feature value range after normalization falls between 0 and 1. The processing unit 140 also normalizes the force feature value to make the normalized force feature value range If it falls between 0 and 1, the processing unit 140 finally inputs the normalized posture feature value and the normalized force feature value into the standard motion model to calculate the posture similarity percentage and the force similarity percentage. The higher the posture similarity percentage or the force similarity percentage, the more similar the user's arm swing action is to the standard arm swing action in the standard motion model. In the disclosed embodiment, if the processing unit 140 first divides an arm swing action into multiple decomposition actions, the processing unit 140 will correspondingly calculate multiple posture similarity percentages and multiple force similarity percentages, which means, The four decomposition actions will correspond to four poses respectively The similarity percentage is the same as the four force similarity percentages.

Please return to Figure 1, the wearable device-based exercise teaching assistance system 100 further includes a man-machine interface 160 for displaying the arm swing action or the posture similarity percentage and force similarity corresponding to each decomposition action through the radar chart interface percentage. FIG. 4 is a schematic diagram of the display screen of the human-machine interface 160 according to the embodiment of the disclosure. As shown in FIG. 4, the human-machine interface 160 includes a posture similarity radar 162 and a force similarity radar 164 for presenting the posture similarity percentage and the force similarity percentage corresponding to each decomposition action through the radar chart interface. As shown in Figure 4, the man-machine interface 160 also includes a teaching suggestion interface 166 for providing action correction suggestions to the user based on the posture similarity percentage and the force similarity percentage corresponding to each decomposition action, so that the user can be specific Know how to perform corrections. In addition, the teaching suggestion interface 166 also includes a coaching instruction input box, so that the coach can write the corresponding instruction based on the posture similarity percentage and strength similarity percentage corresponding to each decomposition action, so that the user can know the need for correction of the action. Pay attention to the details. As shown in Figure 4, the man-machine interface 160 also includes a video menu 168 for providing a coach video menu and a user video menu, so that the user can play the coach video by clicking the coach video menu and/or the user video menu And/or the user's video for action comparison, so that the user can know the difference between his own arm swing and the coach's standard arm swing through the image reproduction.

It is mentioned in the above description that the processing unit 140 will input the normalized posture feature value and the normalized force feature value into the standard motion model to calculate the posture similarity percentage and the force similarity percentage. The establishment process of the standard motion model is explained. In this disclosure In the embodiment, the standard motion model is established by using Back Propagation Artificial Neural Network (BPANN) technology, and standard arm swings made by multiple coaches are used as the training data of the back propagation neural network , Through training to establish a standard action model, the process is as follows. Wear the wearable device 120 on the arm of each of multiple coaches; when the wearable device 120 senses that each coach makes multiple standard arm swings, each standard arm swing corresponds to the three axes at different time points. Angular velocity value and myoelectric signal; the processing unit 140 divides each standard arm swing into multiple standard decomposition actions according to the changes of the three-axis angular velocity value with different time points; the processing unit 140 uses Riemann integration to divide each The three-axis angular velocity values corresponding to each standard decomposition action are converted into posture feature values and normalized; the processing unit 140 converts the myoelectric signal corresponding to each standard decomposition action into force feature values through the average absolute value operation. Normalization; the processing unit 140 uses the normalized posture feature value and the normalized force feature value corresponding to each standard decomposition action as the training data of the backward pass neural network, thereby establishing a standard action through training model.

The division of each standard arm swing into multiple standard decomposition actions described in the above process is basically similar to the aforementioned method of dividing each arm swing into multiple decomposition actions, and will not be repeated here. In the above process, the Riemann integration is used to convert the three-axis angular velocity value corresponding to each standard decomposition action into the posture eigenvalue and the normalization is basically the same as the aforementioned Riemann integration for each decomposition. The three-axis angular velocity value corresponding to the action is converted into the posture feature value and normalized in a similar manner, which will not be repeated here. In the above process, the average absolute value operation is used to divide each standard The EMG signal corresponding to the solution action is converted into a force characteristic value and normalized. It is basically the same as the above-mentioned average absolute value operation to convert the EMG signal corresponding to each decomposition action into a force characteristic value and normalized. The method is similar and will not be repeated here.

In the embodiment of this disclosure, the number of coaches is 4, and each performs 15 standard arm swings, which means that a total of 60 standard arm swings are used as the training data of the backward transfer neural network, but this disclosure Not limited to this. In addition, the purpose of normalizing the posture eigenvalues and the force eigenvalues and then inputting the backward transfer neural network is to speed up the recognition speed of the backward transfer neural network. In the embodiment of the present disclosure, after the standard motion model is established, the processing unit 140 determines the classification accuracy of the standard motion model by 10-fold cross-validation, thereby verifying that the standard motion model is accurate of.

In addition, in the above description of using Figure 3 to explain how the present disclosure divides the arm swing into four decomposition actions, it may happen that the user is not familiar with the standard arm swing and cannot make contact with the coach. Similar arm swing action, as a result, it may not be able to correctly identify the judgment point P4 and/or judgment based on the change of the Y axis angular velocity value among the three-axis angular velocity values corresponding to the arm swing action with different time points Point P8. Therefore, when the start point and/or end point of the decomposition action 3 cannot be determined by the change of the Y-axis angular velocity value with different time points, the processor 140 will use the decomposition corresponding to multiple standard arm swings The average value of point P4 and/or point P8 of action 3 with respect to the time axis is used as the starting point and/or ending point of action 3, and then the action 1 can be found separately according to the aforementioned method. Action 2 The starting point and ending point of the decomposition action 4.

In summary, this disclosure proposes a wearable device-based exercise teaching assistance system, which includes a wearable device and a processing unit. The three-axis gyroscope and the myoelectric sensor included in the wearable device are used to sense the three-axis angular velocity values and myoelectric signals corresponding to different time points when the user wearing the wearable device makes an arm swing. The processing unit receives the three-axis angular velocity value and the EMG signal to calculate the posture feature value and the force feature value corresponding to the arm swing, and inputs the posture feature value and the force feature value into the standard motion model to calculate the posture similarity percentage Percentage of similarity with force. Therefore, with the wearable device-based exercise teaching aid system proposed in this disclosure, the user does not need a coach to watch while training, and the wearable device-based exercise teaching aid system can present the results to inform the user of his arm swing Whether the movement conforms to the standard arm swing movement, which greatly improves the efficiency of individual practice.

The features of several embodiments are summarized above, so those who are familiar with the art can better understand the aspect of the disclosure. Those who are familiar with this technique should understand that they can easily use the present disclosure as a basis to design or modify other processes and structures, thereby achieving the same goals and/or the same advantages as the embodiments described herein. . Those who are familiar with this art should also understand that these equivalent constructions do not depart from the spirit and scope of this disclosure, and they can make various changes, substitutions and alterations without departing from the spirit and scope of this disclosure.

100‧‧‧A sports teaching aid system based on wearable devices

120‧‧‧Wearable device

140‧‧‧Processing unit

160‧‧‧Human Machine Interface

Claims (10)

An exercise teaching assistance system based on a wearable device includes: a wearable device for a user to wear on the arm, the wearable device includes a three-axis gyroscope and at least one electromyography (EMG) sensor The sensors are respectively used to sense a three-axis angular velocity value and at least one myoelectric signal corresponding to different time points when the user makes an arm swing; and a processing unit for receiving the three-axis angular velocity values and the The EMG signal is used to calculate a posture feature value and a force feature value corresponding to the arm swing, and input the posture feature value and the force feature value into a standard motion model to calculate a posture similarity percentage and A force similarity percentage. The wearable device-based exercise teaching aid system described in the first item of the scope of patent application, wherein the processing unit uses Riemann Integral to convert the three-axis angular velocity values into the posture feature values and perform normalization Then, input the normalized posture feature value into the standard motion model to calculate the posture similarity percentage. The wearable device-based exercise teaching assistance system described in the first item of the scope of patent application, wherein the processing unit converts the myoelectric signals into the force characteristic value by means of mean absolute value (Mean Absolute Value) operation. Normalize, and then normalize the characteristics of the force The value is input into the standard motion model to calculate the force similarity percentage. As described in item 1 of the scope of patent application, the wearable device-based exercise teaching assistance system, wherein before the processing unit inputs the posture characteristic value and the force characteristic value into the standard motion model, the processing unit depends on the three-axis angular velocity The value is divided into a plurality of decomposition actions according to the change of the value at different time points, and the processing unit is used to calculate the posture feature value and the strength feature value corresponding to each of the decomposition actions. As described in the first item of the scope of patent application, the exercise teaching assistance system based on the wearable device, wherein the establishment process of the standard action model is as follows: wear the wearable device on the arm of each of the plurality of coaches; The device senses that when each of the coaches makes a plurality of standard arm swings, each of the standard arm swings corresponds to the three-axis angular velocity value and the at least one myoelectric signal at different time points; the processing unit will each The standard arm swing motions are divided into a plurality of standard decomposition motions; the processing unit uses Riemann integration to convert the three-axis angular velocity values corresponding to each of the standard decomposition motions into the posture feature value and perform normalization化; The processing unit converts the myoelectric signals corresponding to each of the standard decomposition actions into the characteristic value of the force by means of the average absolute value operation Line normalization; and the processing unit normalizes the posture eigenvalues and the force eigenvalues corresponding to each of the standard decomposition actions as a Back Propagation Artificial Neural Network (BPANN) The training data to train the standard motion model. According to the wearable device-based exercise teaching assistance system described in item 5 of the scope of the patent application, the processing unit divides each of the standard arm swings into the changes of the three-axis angular velocity value with different time points These standard decomposition actions. The wearable device-based exercise teaching assistance system described in the first item of the scope of patent application, wherein after the establishment of the standard motion model, the processing unit is further determined by 10-fold cross-validation The classification accuracy of the standard motion model. As described in item 4 of the scope of patent application, the wearable device-based exercise teaching assistance system further includes: a human-machine interface for presenting the posture similarity percentage and the corresponding posture similarity of each of the decomposition actions through the radar chart interface The force similarity percentage. As described in item 8 of the scope of patent application, the wearable device-based exercise teaching aid system, wherein the human-machine interface also includes a teaching The suggestion interface is used for providing action correction suggestions to the user according to the posture similarity percentage and the force similarity percentage corresponding to each of the decomposition actions. As described in item 8 of the scope of patent application, the wearable device-based exercise teaching aid system, wherein the man-machine interface also includes a coach video menu and a user video menu, and the user clicks the coach video menu and the User video menu to play video for action comparison.

Images