CN113449945A

CN113449945A - Exercise course scoring method and system

Info

Publication number: CN113449945A
Application number: CN202010228332.0A
Authority: CN
Inventors: 庄龙飞
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-03-27
Filing date: 2020-03-27
Publication date: 2021-09-28
Anticipated expiration: 2040-03-27
Also published as: CN113449945B; WO2021189736A1

Abstract

The invention discloses a sport course grading method and a sport course grading system, wherein the sport course grading method comprises the following steps: playing the course film, obtaining coach movement data corresponding to the course film, and obtaining a coach pane from the coach movement data; acquiring student movement data through an inertia measurement unit, and acquiring a student pane from the student movement data, wherein the length of the student pane is greater than that of the trainer pane; finding a trainee segment in the trainee pane that is most similar to the trainer pane; for the first sport type, calculating a sport score according to the stability of the student segment and the coach pane; and for the second motion type, calculating a motion score with a deduction mechanism and according to the difference between the trainee segment and the trainer pane. Thereby, different scoring methods are employed for different types of sports, whereby a sports score can be calculated more appropriately.

Description

Exercise course scoring method and system

Technical Field

The invention relates to a method and a system for dividing an exercise course into a plurality of exercise types to calculate exercise scores.

Background

With the rise of fitness weathers in recent years, more and more people are willing to try to establish own exercise habits. In the current information-rich environment, many people try to search for a teaching movie related to fitness through resources such as a network and perform exercise by simulating the motion in the movie. However, since it is not known that a general person can correctly perform a desired motion by simulation, the person may be injured by performing a wrong motion before the effect of the exercise is achieved. In addition, without the help of other people, the general user cannot know whether the user gradually progresses in the process of exercise, and therefore the user may not want to continue the exercise because of no achievement.

The current common motion judgment systems include: the motion gesture of the user is captured by utilizing multi-lens photographic equipment and a computer vision processing technology for analysis, and the system has higher cost no matter the hardware cost or the manufacturing cost of digital content, thereby causing difficulty in popularization; one common approach is to use a wearable inertial measurement unit, which at least includes an accelerometer and a magnetometer, to record the acceleration and angular velocity changes of the sensor in three-dimensional space during movement, thereby calculating the movement trajectory in space, which can be recorded and analyzed for comparison. However, different types of exercise have different characteristics, and therefore, how to properly analyze and compare different types of exercise is an issue of concern to those skilled in the art.

Disclosure of Invention

The invention aims to provide a sports course scoring method, which adopts different scoring methods for different sports types, so that the sports score can be calculated more appropriately.

The embodiment of the invention provides a sports course grading method which is suitable for a sports course grading system. The sports course scoring method comprises the following steps: playing the course film, obtaining coach movement data corresponding to the course film, and obtaining a coach pane from the coach movement data; acquiring student movement data through an inertia measurement unit, and acquiring a student pane from the student movement data, wherein the length of the student pane is greater than that of the trainer pane; finding a trainee segment in the trainee pane that is most similar to the trainer pane; for the first sport type, calculating a sport score according to the stability of the student segment and the coach pane; and for the second motion type, calculating a motion score with a deduction mechanism and according to the difference between the trainee segment and the trainer pane.

In some embodiments, the start point of the trainee pane is before the start point of the trainer pane and the end point of the trainee pane is after the end point of the trainer pane. The step of finding the trainee segment in the trainee pane that is most similar to the trainer pane comprises: an open start and end point dynamic time warping algorithm is performed on the trainee pane and the trainer pane to obtain a trainee segment.

In some embodiments, the step of calculating the motion score according to the stability of the student segment comprises: calculating the average absolute error of the student segments on a plurality of sampling points to be used as the stability of the student; calculating the average absolute error of the force path of the training window frame on a plurality of sampling points to be used as the stability of the training; and calculating a stability ratio between the student stability and the trainer stability.

In some embodiments, the step of calculating the motion score according to the stability of the student segment further comprises: converting the stability ratio to a stability score based on the Rogat function; calculating corresponding coach stability for each pane in the coach motion data, and counting a first number of times that the coach stability of the pane is in a first range and a second number of times in a second range; dividing the first order by the sum of the first order and the second order to calculate a stability weight; and calculating the motion score according to the stability score and the stability weight.

In some embodiments, the step of calculating the motion Score according to the stabilization Score and the stabilization weight is calculated according to the following equation (1), wherein Score is the motion Score, S_stabilityTo stabilize the score, w is the stabilization weight, S₁Is a real number.

Score＝S_stability×w+(1-w)×S₁ (1)

In some embodiments, the step of calculating the sport score according to the difference between the trainee segment and the trainer pane by the deduction mechanism comprises: calculating the power error and the direction error between the student segment and the coach pane; setting a power weight and a direction weight, and increasing the corresponding power weight or direction weight when the larger one of the power error and the direction error is larger than an error critical value; and calculating a motion score according to the power track error, the direction error, the power track weight and the direction weight.

In some embodiments, the aforementioned force error is calculated from the difference between the L2 norm of the trainee segment at the paired sample point and the L2 norm of the trainer pane at the corresponding sample point. The directional error is calculated based on the cosine similarity between the paired sample points of the student fragment and the corresponding sample points of the trainer pane.

In some embodiments, the step of calculating the motion Score according to the power error, the direction error, the power weight and the direction weight is calculated according to the following equation (2), wherein Score is the motion Score, and D is_aIs a directional error, w_aIs a directional weight, D_mFor errors of force, w_mAre strength weights.

Score＝100-(D_a×w_a+D_m×w_m) (2)

In another aspect, an embodiment of the present invention provides an exercise course scoring system, which includes a wearable device and an intelligent device. The wearable device comprises an inertia measurement unit for obtaining the movement data of the student. The intelligent device is used for playing the course film through the display and obtaining the coach motion data corresponding to the course film. The intelligent device is used for executing the following steps or transmitting the student motion data and the trainer motion data to the computing module to execute the steps: obtaining a trainee pane from the trainee motion data and a trainer pane from the trainer motion data, wherein the length of the trainee pane is greater than the length of the trainer pane; finding a trainee segment in the trainee pane that is most similar to the trainer pane; for the first sport type, calculating a sport score according to the stability of the student segment and the coach pane; and for the second motion type, calculating a motion score with a deduction mechanism and according to the difference between the trainee segment and the trainer pane.

In the above-described method, different scoring methods are employed for different types of sports, whereby a sports score can be calculated more appropriately.

Drawings

Fig. 1A is a schematic diagram illustrating an athletic lesson scoring system according to an embodiment.

Fig. 1B is a partial flow diagram illustrating the athletic lesson scoring system 100 according to one embodiment.

FIG. 2 is a schematic diagram illustrating trainer motion data and trainee motion data according to one embodiment.

Fig. 3 is a flowchart illustrating a sports lesson scoring method according to an embodiment.

Description of the main reference numerals:

100-sports course scoring system

110-wearable device, 112-student, 120-display, 121-score, 130-intelligent device, 131-processor, 132-memory, 133-wireless communication module, 140-cloud database, 141-calculation module, 151-156-step, 210-trainer movement data, 220-student movement data, 211, 212-trainer pane, 221-student pane, 230-student segment, 231, 232-sampling point, 301-306-step.

Detailed Description

As used herein, "first," "second," …, etc., do not denote any order or sequence, but rather are used to distinguish one element or operation from another element or operation described in the same technical language.

The invention provides an exercise course grading method combining an intelligent device and cloud application service, which can capture exercise data of a student on a body part while the student exercises according to an exercise course film, and compare the exercise data of the student with the exercise data of a coach, so as to evaluate exercise results of a user. Therefore, the trainees can clearly know whether the trainees correctly execute the required actions, and the exercise effect is improved and the exercise will is promoted. In particular, the score may be calculated in different ways for different types of movements, and may be more tolerant for types of movements where the gesture is not important or for fast movements.

Fig. 1A is a schematic diagram illustrating an athletic lesson scoring system according to an embodiment. Referring to fig. 1A, the athletic lesson scoring system 100 includes at least one wearable device 110, a display 120, and an intelligent device 130.

The wearable device 110 may be implemented, for example, as a sports bracelet and worn by the trainee 112, but may also be implemented as a watch, strap, or other device that may be worn on the body in other embodiments. The wearable device 110 includes an inertial measurement unit that includes acceleration sensors to measure acceleration values in three axes, X, Y, Z, and in some embodiments, the inertial measurement unit may further include angular velocity sensors and/or magnetometers. Herein, the data measured by the inertia measurement unit is referred to as trainee exercise data. The wearable device 110 further comprises a wireless communication module, such as a bluetooth communication module, a wireless fidelity (WiFi) module, or other suitable low power wireless transmission module. In some embodiments, the wearable device 110 may further include a display panel or any other elements, which is not limited herein. The display 120 is used to play a course movie in which the trainer is performing a demonstration action.

The intelligent device 130 comprises a processor 131, a memory 132 and a wireless communication module 133, wherein the memory 132 stores program codes and is executed by the processor 131. The processor 131 may be a central processing unit, a microprocessor, a microcontroller, a digital signal processor, an image processing chip, an application specific integrated circuit, etc., but the invention is not limited thereto. The wireless communication module 133 is, for example, a bluetooth communication module, a wireless fidelity (WiFi) module or other suitable low power wireless transmission module, and is used for receiving the trainee exercise data from the wearable device 110.

Fig. 1B is a partial flow diagram illustrating the athletic lesson scoring system 100 according to one embodiment. Referring to fig. 1A and 1B, a course movie is first created, and in step 151, the course movie is shot and the coach exercise data is recorded. In some embodiments, the activities of the trainer can be captured by multiple cameras, which are mounted at different positions and angles, and the captured videos can be edited to generate a course video. In particular, the trainer also wears one or more wearable devices 110 during the shooting process, wherein the signals sensed by the inertia measurement unit are recorded as trainer movement data. For example, the trainer motion data includes a plurality of sample points, each sample point including X, Y, Z, three acceleration values.

In step 152, a post-production of the course video is performed, and the trainer motion data is synchronized with the course video according to the time code. The time code is, for example, a time code specified by Society of Motion Picture and Television Engineers (SMPTE), but in other embodiments, any format of time code may be used, and the invention is not limited thereto. Therefore, it can be known which video or image of the course video corresponds to each sampling point of the coach exercise data through the time code. For example, if the sampling frequency of the inertia measurement unit in the wearable device 110 is 25Hz, the training exercise data corresponding to a 1 minute movie will have 60 × 25 to 1500 sampling points. However, the present invention does not limit how much the sampling frequency is. Next, the class movies and trainer movement data are stored in cloud database 140.

When the user wants to start exercising, the smart device 130 can retrieve the lesson movie and the trainer exercise data from the cloud database 140 and play the lesson movie through the display 120. Meanwhile, in step 153, the intelligent device 130 obtains the trainee exercise data through the wearable device 110 (via the wireless communication module). In some embodiments, the processor 131 may begin receiving student movement data from the wearable device 110 before the session movie begins playing, but the student movement data may be discarded until the session movie begins playing. In this way, the acquired student movement data corresponds to the course movie, for example, it can be known which frame in the course movie each sampling point in the student movement data corresponds to.

At step 154, the acquired trainee exercise data is collected at the end of the exercise break or end (end of the course movie). Since the correspondence between the trainee motion data and the course movie is already known in step 153 and the correspondence between the course movie and the trainer motion data can be obtained according to the SMPTE time code, the trainee motion data and the trainer motion data can be obtained according to the picture number of the course movie.

In step 155, the trainer motion data is divided into a plurality of trainer panes and a motion score is calculated for each trainer pane. Finally, in step 156, the exercise score 121 is displayed, so that the trainee 112 can clearly know whether the trainee performs the required action correctly, thereby improving the exercise effect and increasing the exercise will. In some embodiments, the movement score is calculated by the processor 131, but in other embodiments, the movement score may be calculated by a server on the cloud or other electronic device. For example, in fig. 1A, the smart device 130 may be connected to a computing module 141 on the cloud, and the computing module 141 may be a server, a virtual machine, or a network application providing computing services, but the invention is not limited thereto. The processor 131 can transmit the trainer motion data and the trainee motion data to the calculating module 141, and the calculating module 141 calculates the motion score and then transmits the motion score back to the processor 131. In the embodiment of fig. 1B, the exercise score is calculated after the exercise is stopped or ended, but in some embodiments, the exercise score may be calculated while the trainee exercise data is acquired, that is, the exercise score may be displayed immediately. In some embodiments, the calculation of the score may be started in the intelligent device 130 immediately when the amount of the trainee exercise data is sufficient during the exercise, or the trainee exercise data may be transmitted to the calculation module 141 after the exercise to perform the exercise score calculation. The calculation of the motion score will be described in detail below.

Referring to fig. 2, fig. 2 is a schematic diagram illustrating trainer motion data and trainee motion data according to an embodiment. The trainer motion data 210 and trainee motion data 220 in fig. 2 are illustrated as one-dimensional signals (i.e. only one acceleration value is provided at each sampling point), but this is only a schematic diagram, and actually each sampling point should have a plurality of acceleration values (i.e. constitute a vector), and the trainer motion data 210 and trainee motion data 220 are a set of vectors. Specifically, coaching motion data 210 can be expressed as

Wherein i is a positive integer representing the ith sampling point,

respectively, the acceleration values of the trainer motion data 210 in the X-axis, Y-axis, and Z-axis at the ith sampling point. In addition, the trainee motion data 220 is represented as

Wherein

Which respectively represent the acceleration values of the trainee's motion data 220 in the X-axis, Y-axis and Z-axis at the ith sampling point. In fig. 2, the coaching motion data 210 and the trainee motion data 220 correspond to the same lesson movie, so if the coaching motion data 210 and the trainee motion data 220 are more similar, the calculated score should be higher. Notably, the length of the trainer motion data 210 may be different than the length of the trainee motion data 220. In some embodiments, if the trainee motion data 220 is sampled at a different frequency than the trainer motion data 210, the one with the higher sampling frequency may be re-sampled (sampling), thereby making the sampling frequency of the two the same.

In some embodiments, if the inertial measurement unit further includes an angular accelerometer, a six-axis sensor data fusion algorithm may be used to obtain a three-dimensional acceleration vector F_c[i]、F_s[i]Therefore, the gesture and the direction of the motion can be distinguished properly, more various motion types can be judged, and better discrimination rate is provided. If the inertia measurement unit further comprises a three-axis magnetometer, the three-dimensional acceleration vector F can be obtained more accurately by using an algorithm of data fusion of a nine-axis sensor_c[i]、F_s[i]So as to perform faster comparison and calculate the similarity rate with accuracy. The fusion algorithm may use any AHRS (attitude and heading reference system) algorithm, but the invention is not limited thereto.

First, the trainer pane 211 is obtained from the trainer motion data 210, and the length of the trainer pane 211 can be determined according to different motion types, and can be 2 seconds to 10 seconds, which is not limited in the present invention. In addition, a trainee pane 221 is also taken from the trainee motion data 220, wherein the trainee pane 221 will be longer than the trainer pane 211. For example, the start of the trainee pane 221 may be before the start of the trainer pane 211 and the end of the trainee pane 221 may be after the end of the trainer pane 211, because the trainee may experience a delay or an early event when mimicking the action of a trainer, and a longer trainee pane 221 may be more tolerant of these events. Assuming that the length of the training pane 211 is 4 seconds and the sampling frequency is 25Hz, the training pane 211 has a total of 4 × 25 — 100 sampling points, which are denoted as positive integer M below. If the trainee pane 221 extends 1 second forward and 1 second backward from the coach pane 211 and the sampling frequency of the trainee pane 221 is also 25Hz, the trainee pane 221 has a total of 100+2 × 25 — 150 sampling points, which are denoted as a positive integer N below.

Next, the trainee segment 230 most similar to the trainer pane 211 is found in the trainee pane 221, and in some embodiments, an open-start-end dynamic time warping (OBE-DTW) algorithm is performed on the trainee pane 221 and the trainer pane 211 to obtain the trainee segment 230. Those skilled in the art can understand that the general dynamic time warping algorithm is not described herein, and OBE-DTW releases the same constraint of starting point and ending point, so as to find the segment of partial matching from the learner pane 221, which can be expressed as the following equations (1) - (3).

Wherein

The subset in the trainee pane 221 is composed of the sample points from the d-th to the e-th, wherein d and e are positive integers, and the segments

May be the same length as the trainee pane 221 or may be different. The algorithm is to find the positive integers d, e from the learner pane 221, so that the error is caused

Will be minimal, where Cost_OBE-DTWRepresenting the average of the error between the paired sample points according to OBE-DTW, and the paired sample points in the trainee pane 221 constitute the trainee segment 230. For example, the sampling point 231 in the trainee segment 230 is paired to the sampling point 232 in the trainer pane 211, and it is noted that the sampling point in the trainee segment 230 may be repeated (paired to multiple sampling points in the trainer pane 211) and not necessarily continuous (some sampling points are not paired to the trainer pane 211). In some embodiments, different errors may be used depending on the type, and if the evaluation is motion-direction-dominant, Cost_OBE-DTWRefers to the average of cosine similarity (cosine similarity), Cost if it is an assessment dominated by motion power_OBE-DTWRefers to the average of the norm of L2 (also known as the euler distance). Furthermore, N_matctedWhich refers to the length of the student fragment found and a is a real number, can be determined experimentally (e.g., 2). The above equation (3) is to adjust the error according to the length of the student segment

The length of the student fragment 230 found may be smaller or larger than the length of the trainer pane 211, so that the length of a fragment is preferably selected as the length of the trainer pane 211 is closer, unlike the conventional OBE-DTW. As shown in fig. 2As shown, the student's movements are slightly behind the trainer's movements, so the first portion of the student pane 221 is discarded according to the above algorithm, matching only the second half of the student pane 221, which gives the student a greater tolerance for the movements behind.

After finding the trainee segment 230, the trainee segment 230 is analyzed for the type of exercise, and different types of exercise may take different score calculations. At least two motion types are defined, a first motion type being a stable static motion and a second motion type being a general or fast dynamic motion. In some embodiments, the absolute values of the acceleration values in the X, Y, and Z axes at each sampling point may be added to obtain a force channel, which is expressed by the following equation (4).

x[i]＝|a_i,x|+|a_i,y|+|a_i,z| (4)

The acceleration values in equation (4) may pertain to trainer motion data or trainee motion data. When the type of exercise is to be determined, the power of the exercise data can be used, and then the classification of the exercise is performed according to the statistical values such as the average, standard deviation or mean absolute difference (mean absolute difference) of all power x [ i ], for example, when the mean absolute difference is less than a threshold value, it can be determined as a first type of exercise, and when the mean absolute difference is greater than or equal to the threshold value, it can be determined as a second type of exercise. In some embodiments, the trainer motion data may also be categorized according to the sport item to which the trainer motion data pertains (e.g., boxing aerobics, latin dances, intermittent intensity exercises, yoga, etc.), or by the length of the trainer pane 211 (or trainee pane 221). In some embodiments, the trainee pane 221 may also be weighted for the first motion type (between 0-1) and the second motion type (between 0-1), and the scores for the first motion type and the second motion type are computed separately, and then the scores are summed according to the weights, and different trainee panes 221 may belong to different motion types or have different motion type weights. The invention does not limit how the motion classification is performed. The calculation of the motion score is described below in two motion types.

[ first type of motion ]

The above-mentioned average absolute deviation can be expressed as the following equation (5), and the average absolute deviation is referred to herein as the stability MAD.

Where mu is all force paths x [ i ]]Is calculated. Equation (5) is used to calculate the stability MAD of the trainee's motion data, hereinafter called the trainee stability MAD_user. Equation (5) can also be used to calculate the stability of the trainer motion data by replacing the positive integer N in equation (5) with M, which is referred to as trainer stability MAD_ref. Then the stability of the student is determined_userStability to coach MAD_refThe more similar the two are, the higher the calculated motion score. In some embodiments, a student stability MAD may be calculated_userStability to coach MAD_refThe stability ratio r therebetween is shown in the following equation (6).

The stability ratio is then shifted, scaled, and flipped based on the rogit function (logit function), thereby converting to a stability score, as shown in the following equations (7) to (9).

Wherein S_stability(r) represents the stability score. S_baseIs a real number, which represents when r is 1, i.e. the student stability MAD_userStability to coach MAD_refThe same may be said of the scores. r is_boundIs a real number, defining a threshold value for the stability ratio r, when set to r_boundA score of 100 is obtained at 5, which corresponds to a trainee stability of five times the trainer stability (r-1/5), and a score of zero at one fifth of the trainer stability (r-5). q is a real number used to control the variation of equation (8), and can be obtained through experiments. In some embodiments, the above-described Rogat function may be implemented as a look-up table.

In order to solve the problem that the scoring condition is too strict in response to the diversity of the motion context, which may cause many visually static motions to be classified as dynamic motions and cause too many misjudgments, the embodiment uses a dual threshold to subdivide the static motion into true static motion and fuzzy static motion. In particular, two thresholds are defined here, respectively an upper static threshold limit t_hAnd a true static threshold t, wherein the static threshold is upper bound t_hGreater than the true static threshold t, which determines two ranges, respectively the first range 0, t]And a second range (t, t)_h). Next, the coach motion data is divided into a plurality of panes, and for each pane, a corresponding coach stability MAD is calculated_refAnd counting these coach stability MADs_refIn a first range [0, t]First degree and second range (t, t)_h) And then divides the first number by the sum of the first number and the second number to calculate a stabilization weight w, which is expressed as the following equation (10).

Wherein count_underCount as the first number of times_fuzzyThe second number. Finally, according to the stable score S_stability(r) and the stabilization weight w, as shown in equation (11) below.

Score＝S_stability×w+(1-w)×S₁ (11)

Where Score is the motor Score. S₁A real number can be set experimentally, such as 90.

[ second motion type ]

This sport type is a scoring mechanism and calculates a sport score based on the difference between the trainee segment and the trainer pane. First, the power error between the student segment and the trainer pane is calculated, and the power error is shifted and scaled by the S function (sigmoid function). Specifically, the S function is shown in the following equation (12).

Wherein m and q are parameters. In some embodiments, the S function may be implemented as a look-up table. The power of each sample point in the coach pane can be expressed as equation (13) below.

mag_c[i]＝‖F_c[i]‖,i＝1,2,…M (13)

Where | represents the L2 norm, M is the length of the coach pane. The dynamics factor of the trainer motion data is then calculated as shown in equation (14) below.

mag_upper[i]＝std(mag_c,i,w)+mean(mag_c,i,w),

i＝1,2,…,length(mag_c) (14)

Where w represents the pane size, std (mag)_cI, w) represents force mag_c[]Standard deviation, mean (mag) between the ith to the i + w th sample point_cI, w) represents force mag_c[]Average value from the i-th sampling point to the i + w-th sampling point. Then, the S function is transformed as shown in the following equation (15).

Wherein P represents a set comprisingThe paired sample points between the student fragment and the coach pane are generated by the OBE-DTW algorithm, such as sample point 231 in FIG. 2 is paired to sample point 232, (i)_c,i_s) Representing each pair in the set P. m is_m、q_mThe real number can be arbitrarily set by experiment to control the tolerance of the force difference. D_mThe score to be deducted is indicated, i.e. the above-mentioned force error. Dynamic factor mag_upper[i_c]Is used for adaptively adjusting the power track error D according to the coach motion data_mWhen the standard deviation of the power of the trainer movement data is large, the action is difficult to imitate, so the deducted score D_mThere will be less. In another aspect, the force error D_mIs based on the L2 norm | F of the student segment at each sampling point_s[i_s]II F h L2 norm on corresponding sample point with coach pane_c[i]The difference between |.

In addition, the angle difference is calculated according to the following equation (16).

Wherein m is_a、q_aThe tolerance for the angle difference can be arbitrarily set by experiment. D_aThe fraction to be deducted, i.e. the above-mentioned directional error, is indicated. In another angle, the directional error D_aIs based on each sampling point F of the student's segment_s[i_s]Corresponding sample point F with coach pane_c[i_c]The cosine similarity between them. In other words, the force error D_mIs used to judge whether the force of the student is correct or not and the direction error D_aIs used to determine whether the direction of the student's action is correct. When the student follows the coach, the difference between the power way and the direction is not too large, and the point is deducted little. However, when the student and the trainer have great difference in movement, there is a possibility that the student and the trainer have great similarity in direction or power, and a high score should not be given at the moment, so that the student and the trainer need to have great difference in movementTwo weights are set to blend the errors D_mError from direction D_a。

Specifically, the weight w is also set in this embodiment_mAnd a direction weight w_aWhen the larger of the force error and the direction error is larger than the error threshold (indicating that the student and the trainer have great difference in movement), the corresponding force weight or direction weight is increased, and the other weight is decreased, so that the user does not give a high score due to the fact that the user has great similarity in direction or force. In some embodiments, the step of adding weight is adjusted according to an S function, as shown in equation (17) below.

Furthermore, the weight w_mAnd a direction weight w_aIs 1, weight w_mAnd a direction weight w_aIs determined according to the following virtual code.

Wherein q is a real number for determining the magnitude of the weight increase. m is a real number, which is used to determine the displacement of the S function on the X-axis, i.e. representing the threshold value. The real numbers q, m can be arbitrarily set through experiments. In other words, when the direction error D is_aWhen large, the direction error D_aWill be input to the S function when the direction error D_aIf the value is larger than the critical value, the value will rise rapidly, and the change amplitude of other conditions is very low.

Finally, a motion score is calculated based on the power error, the direction error, the power weight, and the direction weight, as shown in equation (18) below.

Score＝100-(D_a×w_a+D_m×w_m) (18)

Referring back to fig. 2, after the motion score is calculated for the trainer pane 211, the motion score of the next trainer pane 212 can be calculated, particularly with the trainer pane 212 partially overlapping the trainer pane 211. For example, the trainer pane 211 can be slid to the right a distance to get the trainer pane 212, and then the motion score of the trainer pane 212 can be calculated. Too small a sliding distance increases many useless calculations and does not improve the resolution of the motion analysis, too large a sliding distance decreases many calculations but also decreases the resolution of the motion analysis, which in some practical applications is 0.15 seconds.

Fig. 3 is a flowchart illustrating a sports lesson scoring method according to an embodiment. Referring to fig. 3, in step 301, a course movie is played, coach exercise data corresponding to the course movie is obtained, and a coach pane is obtained from the coach exercise data. In step 302, trainee movement data is obtained via the inertial measurement unit, and a trainee pane is obtained from the trainee movement data, wherein the trainee pane has a length greater than that of the trainer pane. In step 303, the trainee segment in the trainee pane that is most similar to the trainer pane is found. In step 304, the type of movement is determined, and if the type of movement is a first type of movement, in step 305, a movement score is calculated based on the stability of the trainee segment and the trainer pane, and if the type of movement is a second type of movement, in step 306, a movement score is calculated based on the difference between the trainee segment and the trainer pane with a point deduction mechanism. In another embodiment, step 304 may be replaced by calculating the weight of the first motion type and the weight of the second motion type, and steps 305 and 306 are performed, and then the motion scores calculated by

steps

305, 306 are added according to the weight of the first motion type and the weight of the second motion type. As described above, the weights of the first exercise type and the second exercise type may be determined by a table lookup according to an exercise item (e.g., boxing oxygen, latin dance, intermittent strength exercise, yoga, etc.), or may be determined according to the length of the trainer pane 211 (or the trainee pane 221). However, the steps in fig. 3 have been described in detail above, and are not described again here. It is to be noted that, the steps in fig. 3 can be implemented as a plurality of program codes or circuits, and the invention is not limited thereto. In addition, the method of fig. 3 can be used with the above embodiments, or can be used alone. In other words, other steps may be added between the steps of fig. 3.

In some embodiments, the L2 norm may be replaced by an L1 norm, a maximum norm, a histogram distance (histogram distance), or other suitable distance calculation method. In some embodiments, the OBE-DTW may be replaced with other DTWs, or with a Longest Common Subsequence (LCS) or other similarity algorithm.

Although the present invention has been described with reference to the above embodiments, it should be understood that various changes and modifications can be made therein by those skilled in the art without departing from the spirit and scope of the invention.

Claims

1. A sports lesson scoring method is suitable for a sports lesson scoring system, wherein the sports lesson scoring system comprises a wearable device, the wearable device comprises an inertia measurement unit, and the sports lesson scoring method comprises the following steps:

playing the course video, obtaining coach movement data corresponding to the course video, and obtaining a coach pane from the coach movement data;

obtaining trainee movement data through the inertia measurement unit, and obtaining a trainee pane from the trainee movement data, wherein the length of the trainee pane is greater than that of the coach pane;

finding a trainee segment in a trainee pane that is most similar to the trainer pane;

for a first type of motion, calculating a motion score according to the stability of the trainee segment and the trainer pane; and

for a second type of motion, calculating the motion score with a deduction mechanism and according to the difference between the trainee segment and the trainer pane.

2. The athletic lesson scoring method of claim 1, wherein the starting point of the trainee pane is before the starting point of the trainer pane and the ending point of the trainee pane is after the ending point of the trainer pane, and wherein the step of finding the trainee segment in the trainee pane that is most similar to the trainer pane comprises:

an open start-end dynamic time warping algorithm is performed on the trainee pane and the trainer pane to obtain the trainee segment.

3. The athletic lesson scoring method of claim 1, wherein the step of calculating the athletic score based on the stability of the student segment comprises:

calculating the average absolute error of the student segments on a plurality of sampling points to serve as the stability of the student;

calculating the average absolute error of the force path of the training window pane on a plurality of sampling points to be used as training stability; and

and calculating a stability ratio between the student stability and the coach stability.

4. The athletic lesson scoring method of claim 3, wherein the step of calculating the athletic score based on the stability of the student segment further comprises:

converting the stability ratio to a stability score based on a Rogott function;

calculating the corresponding trainer stability for each pane in the trainer motion data, and counting a first number of times that the plurality of trainer stabilities of the plurality of panes are in a first range and a second number of times in a second range;

dividing the first number of times by a sum of the first number of times and the second number of times to calculate a stabilization weight; and

calculating the motion score according to the stabilization score and the stabilization weight.

5. The exercise lesson scoring method of claim 4, wherein the step of calculating the exercise score based on the stability score and the stability weight is calculated according to the following equation (1):

Score＝S_stability×w+(1-w)×S₁ (1)

wherein Score is the motion Score, S_stabilityIs the stability score, w is the stability weight, S₁Is a real number.

6. The athletic lesson scoring method of claim 1, wherein the step of calculating the athletic score based on the difference between the trainee segment and the trainer pane in a scoring mechanism comprises:

calculating a power and direction error between the trainee segment and the trainer pane;

setting a power weight and a direction weight, and increasing the corresponding power weight or the direction weight when the larger one of the power error and the direction error is larger than an error critical value; and

and calculating the motion score according to the power error, the direction error, the power weight and the direction weight.

7. The athletic lesson scoring method of claim 6, wherein the moral error is calculated as a difference between an L2 norm of the student's segment at a paired sample point and an L2 norm of the trainer pane at a corresponding sample point,

wherein the directional error is calculated from a cosine similarity of the trainee segment between the paired sample point and the corresponding sample point of the trainer pane.

8. The athletic lesson scoring method of claim 7, wherein the step of calculating the athletic score based on the power error, the directional error, the power weight, and the directional weight is calculated according to the following equation (1):

Score＝100-(D_a×w_a+D_m×w_m) (1)

wherein Score isThe movement score, D_aFor the directional error, w_aAs the directional weight, D_mFor the force error, w_mIs the force path weight.

9. An athletic lesson scoring system, comprising:

the wearable device comprises an inertia measurement unit, a data acquisition unit and a data processing unit, wherein the inertia measurement unit is used for acquiring the movement data of a student; and

the intelligent device is used for playing the course film through the display and obtaining the coach motion data corresponding to the course film;

wherein the intelligent device is configured to perform the following steps or transmit the trainee motion data and the trainee motion data to a computing module to perform the steps:

obtaining a trainee pane from the trainee motion data and a trainer pane from the trainer motion data, wherein the trainee pane has a length greater than a length of the trainer pane;

finding a trainee segment in the trainee pane that is most similar to the trainer pane;