Peking opera performance action evaluation method and device based on action capture data

Abstract

The invention provides a motion capture data-based Beijing opera performance action evaluation method and apparatus, and relates to the technical field of somatosensory interaction, the motion capture data-based Beijing opera performance action evaluation method comprises the steps of obtaining motion capture data of a to-be-evaluated Beijing opera performance action; inputting the motion capture data into a motion evaluation model, and outputting a similarity score of the Beijing opera performance motion to be evaluated; wherein the action evaluation model is obtained by training on the basis of samples of Beijing opera performance actions and corresponding similarity scores, the similarity scores are predetermined according to the samples of the Beijing opera performance actions and are in one-to-one correspondence with the samples of the Beijing opera performance actions, and the defects of differences between the actions can be quantified; peking opera performance actions are assisted to be mastered.

Description

Technical Field

[0001] The present invention relates to the field of somatosensory interaction technology, and in particular to a method and device for evaluating Peking Opera performance movements based on motion capture data. Background Art

[0002] Peking Opera, as one of my country's cultural treasures and quintessence, has profound artistic connotations and unique performance forms, and is an important part of China's excellent traditional culture. In the process of teaching traditional Peking Opera performance movements, teachers usually adopt the method of "oral transmission and heart-to-heart teaching", relying on years of performance experience and the "feeling" of art to provide guidance. There are few precise movement specifications and quantitative movement indicators. For young learners who lack performance experience, it is often difficult to feel the difference between their own movements and those of teachers, and they cannot accurately grasp the shortcomings of their own performance movements. This has brought considerable difficulties to the current teaching of Peking Opera performance movements and the inheritance of Peking Opera art. Summary of the invention

[0003] The present invention provides a method and device for evaluating Peking Opera performance movements based on motion capture data, so as to solve the defect that it is difficult to quantify the differences between movements in the prior art and to assist in grasping the shortcomings of Peking Opera performance movements.

[0004] The present invention provides a method for evaluating Peking Opera performance movements based on motion capture data, comprising: obtaining motion capture data of Peking Opera performance movements to be evaluated; Inputting the motion capture data into a motion evaluation model, and outputting a similarity score of the Peking Opera performance motion to be evaluated; The action evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores. The similarity scores are predetermined based on the samples of Peking Opera performance actions and correspond one-to-one with the samples of Peking Opera performance actions.

[0005] According to a method for evaluating Peking Opera performance movements based on motion capture data provided by the present invention, the movement evaluation model comprises: a static movement evaluation unit, a dynamic movement evaluation unit and an output unit; The training method of the action evaluation model includes: Processing the sample of the Peking Opera performance action through the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and outputting the predicted similarity score of the Peking Opera performance action; The predicted similarity score is compared with the similarity score, and the parameters of the output unit are adjusted according to the comparison result, until the number of iterations reaches a threshold or the fitness value of the predicted similarity score and the similarity score is less than a threshold.

[0006] According to a method for evaluating Peking Opera performance actions based on motion capture data provided by the present invention, samples of the Peking Opera performance actions are processed by the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and a predicted similarity score of the Peking Opera performance actions is output, including: Inputting the sample of the Peking Opera performance action into the static action evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static action similarity score of the corresponding non-terminal joint points of the Peking Opera performance action and the pre-stored corresponding reference action; Sending the sample of the Peking Opera performance action and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The samples of the Peking Opera performance movements and the dynamic movement similarity scores are sent to the output unit to obtain the predicted similarity scores of the Peking Opera performance movements.

[0007] According to a Peking Opera performance action evaluation method based on motion capture data provided by the present invention, the sample of the Peking Opera performance action and the dynamic action similarity score are sent to the output unit to obtain the predicted similarity score of the Peking Opera performance action, including: Sending the sample of the Peking Opera performance action and the dynamic action similarity score to the output unit to determine the change activity of each non-terminal joint point of the Peking Opera performance action; According to the change activity of each non-terminal joint point, the dynamic action similarity score of each non-terminal joint point is assigned corresponding weights, and the predicted similarity score of Peking Opera performance actions is obtained by adding them up.

[0008] According to a Peking Opera performance action evaluation method based on motion capture data provided by the present invention, the predicted similarity score is compared with the similarity score, and the parameters of the output unit are adjusted according to the comparison result, including: The predicted similarity score of each Peking Opera performance action is compared with the similarity score of the corresponding Peking Opera performance action to obtain the difference; According to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, the average value of the difference between the predicted similarity score of all Peking Opera performance actions and the similarity score of the corresponding Peking Opera performance action is obtained as the fitness value of the predicted similarity score and the similarity score; The parameters of the output unit are adjusted according to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, until the number of iterations reaches a threshold or the average value of the difference between the predicted similarity score of the Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action is less than a threshold.

[0009] According to a method for evaluating Peking Opera performance movements based on motion capture data provided by the present invention, the static movement evaluation unit and the dynamic movement evaluation unit respectively include non-terminal joint angle data of the reference movement; Inputting the motion capture data into the action evaluation model and outputting the similarity score of the Peking Opera performance action to be evaluated includes: Inputting the motion capture data into the static motion evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action to be evaluated and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static motion similarity score of the non-terminal joint points of the Peking Opera performance action to be evaluated and the pre-stored corresponding reference action; Sending the motion capture data and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The motion capture data and the dynamic action similarity score are sent to the output unit, and the similarity score of the Peking Opera performance action to be evaluated is output.

[0010] The present invention also provides a Peking Opera performance action evaluation device based on motion capture data, comprising: A motion capture data acquisition module, used to acquire motion capture data of the Peking Opera performance movements to be evaluated; A similarity score acquisition module is used to input the motion capture data into an action evaluation model and output a similarity score of the Peking Opera performance action to be evaluated; wherein the action evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions and corresponds one-to-one to the samples of the Peking Opera performance actions.

[0011] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for evaluating Peking Opera performance movements based on motion capture data as described above is implemented.

[0012] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for evaluating Peking Opera performance movements based on motion capture data as described above is implemented.

[0013] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for evaluating Peking Opera performance movements based on motion capture data.

[0014] The Peking Opera performance action evaluation method and device based on motion capture data provided by the present invention can fully utilize the motion capture data of the Peking Opera performance action to be evaluated through the action evaluation model, judge the similarity between the Peking Opera performance action to be evaluated and the reference action, thereby quantifying the difference between the Peking Opera performance action to be evaluated and the reference action, and assisting in grasping the shortcomings of the Peking Opera performance action. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flow chart of a method for evaluating Peking Opera performance movements based on motion capture data provided by the present invention.

[0017] Figure 2 It is a schematic diagram of the structure of a Peking Opera performance action evaluation device based on motion capture data provided by the present invention.

[0018] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Combine the following Figure 1-Figure 2 The present invention describes a method and device for evaluating Peking Opera performance movements based on motion capture data.

[0021] Figure 1FIG. 1 is a flow chart of a method for evaluating Peking Opera performance movements based on motion capture data provided by the present invention. Figure 1 As shown, the method includes: Step 101: Obtain motion capture data of the Peking Opera performance movements to be evaluated.

[0022] Motion capture data is motion data obtained using motion capture technology. Motion capture technology (Motion Capture Technology) is a technology used to record and analyze the movements of the human body or other organisms. It can capture motion data by attaching sensors or markers to the performer, and generate accurate motion data by recording the movement trajectory of the sensor or marker in space over time.

[0023] Step 102: input the motion capture data into a motion evaluation model, and output a similarity score of the Peking Opera performance action to be evaluated; wherein the motion evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions and corresponds one-to-one to the samples of the Peking Opera performance actions.

[0024] Specifically, the similarity score of the sample of Peking Opera performance action is a score indicating the similarity between the sample of Peking Opera performance action and the reference action. The reference action may be a standard action of Peking Opera performance action, such as a teacher's Peking Opera performance action, or an expert's Peking Opera performance action. Taking the reference action as a teacher's Peking Opera performance action as an example, different teachers may have different characteristics when performing the same Peking Opera performance action, that is, the Peking Opera performance actions of different teachers may also be different. The reference action here is a Peking Opera performance action of a teacher selected as a standard, or a fixed Peking Opera performance action is used as a reference action.

[0025] The similarity score of the sample of Peking Opera performance action can be given by Peking Opera experts for each sample of Peking Opera performance action. Exemplarily, Peking Opera experts can give an overall rating grade for the sample of Peking Opera performance action, and the overall rating grade may include excellent, good, passing and need to work hard. The initial similarity score of the sample of Peking Opera performance action can be obtained through the initial action evaluation model. Specifically, taking a sample of Peking Opera performance action of a sub-action of a "starting to dominate" action as an example, the initial similarity score of multiple sub-actions of the "starting to dominate" action with an overall rating grade of excellent is obtained through the initial action evaluation model, and the initial similarity score of each sub-action of the "starting to dominate" action may be different. The median of the initial similarity scores of the multiple sub-actions of the "starting to dominate" action is obtained as the similarity score of the sub-action of the "starting to dominate" action with an overall rating grade of excellent.

[0026] Before executing step 102, a motion evaluation model can be pre-trained. Specifically, the motion evaluation model can be trained in the following manner: first, a large number of samples of Peking Opera performance movements are collected, and similarity scores of the samples of Peking Opera performance movements are obtained by combining expert scores with the initial motion evaluation model. The samples of Peking Opera performance movements and their similarity scores are input into the initial motion evaluation model for training, thereby obtaining the motion evaluation model.

[0027] In the embodiment of the present invention, the motion evaluation model can fully utilize the motion capture data of the Peking Opera performance action to be evaluated, judge the similarity between the Peking Opera performance action to be evaluated and the reference action, thereby quantifying the difference between the Peking Opera performance action to be evaluated and the reference action, and assisting in grasping the shortcomings of the Peking Opera performance action.

[0028] Based on the above embodiment, the action evaluation model includes: a static action evaluation unit, a dynamic action evaluation unit and an output unit; The training method of the action evaluation model includes: Processing the sample of the Peking Opera performance action through the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and outputting the predicted similarity score of the Peking Opera performance action; The predicted similarity score is compared with the similarity score, and the parameters of the output unit are adjusted according to the comparison result, until the number of iterations reaches a threshold or the fitness value of the predicted similarity score and the similarity score is less than a threshold.

[0029] The static action evaluation unit can calculate the static action similarity of the samples of Peking Opera performance actions and the corresponding pre-stored reference actions. However, for the same Peking Opera performance action, it may be impossible for two people to perform the same action at the same time due to the differences in performance rhythm, habits, on-site conditions, etc. between the performers of the Peking Opera performance action samples and the performers of the reference actions. The action sequence length of the Peking Opera performance action samples is inconsistent with the action sequence length of the reference actions, and the action trend amplitude changes of the two are different. The dynamic action evaluation unit can reduce the similarity result error caused by this (the specific principle can be seen in the relevant description below).

[0030] When evaluating Peking Opera performance movements, Peking Opera experts pay different attention to different parts, which directly affects the experts' evaluation and is reflected in the experts' overall score. The corresponding output unit can adjust the parameters of the output unit to assign different weights to different parts of the Peking Opera performance movement, so that the predicted similarity score of the output Peking Opera performance movement fits the similarity score of the Peking Opera performance movement.

[0031] The parameters of the output unit are iteratively adjusted according to the comparison results, which means that the weights of different parts of the Peking Opera performance movements are iteratively updated so that the predicted similarity score of the output Peking Opera performance movements gradually approaches the similarity score of the Peking Opera performance movements. Therefore, when the number of iterations reaches a threshold, that is, after a certain number of iterations is reached, the predicted similarity score of the output Peking Opera performance movements can be made basically the same as the similarity score of the Peking Opera performance movements.

[0032] The fitness value of the predicted similarity score and the similarity score can be the average of the difference between the predicted similarity scores of all Peking Opera performance actions and the similarity scores of the corresponding Peking Opera performance actions. In this way, the lower the fitness value, the closer the predicted similarity score is to the similarity score. Therefore, the fitness value of the predicted similarity score and the similarity score is less than the threshold value, which can make the predicted similarity score of the output Peking Opera performance action and the similarity score of the Peking Opera performance action basically the same.

[0033] The specific threshold of the number of iterations and the threshold of the fitness value of the predicted similarity score and the similarity score can be determined according to the specific application scenario and requirements. For example, if a more accurate similarity judgment is required, the threshold of the number of iterations can be increased, and the threshold of the fitness value of the predicted similarity score and the similarity score can be reduced.

[0034] Specifically, comparing the predicted similarity score with the similarity score, and adjusting the parameters of the output unit according to the comparison result may include: The predicted similarity score of each Peking Opera performance action is compared with the similarity score of the corresponding Peking Opera performance action to obtain the difference; According to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, the average value of the difference between the predicted similarity score of all Peking Opera performance actions and the similarity score of the corresponding Peking Opera performance action is obtained as the fitness value of the predicted similarity score and the similarity score; The parameters of the output unit are adjusted according to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, until the number of iterations reaches a threshold or the average value of the difference between the predicted similarity score of the Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action is less than a threshold.

[0035] A Peking Opera performance action may include multiple sub-actions. Take the Peking Opera performance action "Qi Ba" as an example, where each Peking Opera performance action is a sub-action in the Peking Opera performance action "Qi Ba", and all Peking Opera performance actions are the Peking Opera performance action "Qi Ba". By calculating the difference between the predicted similarity score of the current sub-action and the similarity score of the corresponding current sub-action as the "closeness value", accumulating the "closeness value" of each sub-action and averaging it with the number of sub-actions included in the Peking Opera performance action "Qi Ba", the overall fit degree of the Peking Opera performance action "Qi Ba" under the parameters of the current output unit is obtained. Based on the overall fit degree, it is natural to judge whether the parameters of the output unit are appropriate.

[0036] Exemplarily, a particle swarm optimization algorithm may be applied to adjust the parameters of the output unit according to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action (the specific principle may be found in the relevant description below).

[0037] Based on any of the above embodiments, the sample of the Peking Opera performance action is processed by the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and the predicted similarity score of the Peking Opera performance action is output, including: Inputting the sample of the Peking Opera performance action into the static action evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static action similarity score of the corresponding non-terminal joint points of the Peking Opera performance action and the pre-stored corresponding reference action; Sending the sample of the Peking Opera performance action and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The samples of the Peking Opera performance movements and the dynamic movement similarity scores are sent to the output unit to obtain the predicted similarity scores of the Peking Opera performance movements.

[0038] The following is a description of the static action evaluation unit processing a sample of Peking Opera performance actions.

[0039] The sample representation of the Peking Opera performance action is a motion capture data of the Peking Opera performance action as a sample, which may include three-dimensional coordinate data with a time series, and these coordinate data are represented in the form of a three-dimensional floating point vector. Based on these coordinate data, the world coordinate position of each non-terminal joint point in the three-dimensional space can be calculated. However, the world coordinate position is not suitable for direct action calculation and analysis, because the performance and movement of the actors in the venue may cause deviations in the calculation results. Therefore, in this embodiment, the local coordinate position of each non-terminal joint point relative to the skeleton reference point (i.e., the Hip non-terminal joint point) is used, and then the non-terminal joint point angle data of the Peking Opera performance action can be obtained by calculating the Euclidean distance. The non-terminal joint point angle is the angle between the two segments of the skeleton connected by the non-terminal joint point.

[0040] Taking the angle data of the right knee joint point of a Peking Opera performance as an example, when calculating the angle of the right knee joint point of a Peking Opera performance action, the three-dimensional coordinates (x a ,y a , z a ), the three-dimensional coordinates (x b ,y b , z b ) and the three-dimensional coordinates (x c ,y c , z c ), then calculate the vector data of the right knee joint point and the parent node, = (x c -x a ,y c -y a , z c -z a ), calculate the vector data of the right knee joint and child nodes = ( x c -x b , y c -y b , z c -z b ), finally, the angle of the right knee joint can be calculated by the following formula:

[0041] Among them, the parent node is the other end joint point of one bone of the two bones connected by the right knee joint point, and the child node is the other end joint point of one bone of the other bone of the two bones connected by the right knee joint point.

[0042] Based on a similar method, the angle feature of the right knee joint of the corresponding reference action can be obtained in advance. Then, for the right knee joint, the angle difference between the right knee joint angle of the Peking Opera performance action and the right knee joint angle of the reference action is , which can be expressed as the ratio of the absolute value of the difference between the right knee joint angle of the Peking Opera performance action and the right knee joint angle of the reference action to the right knee joint angle of the reference action, that is:

[0043] Based on a similar method, the angle differences of other joints of Peking Opera performances can be obtained, which can be expressed as ∆θ1, ∆θ2, ∆θ3, ..., ∆θn. Finally, the cumulative sum S of the angle differences of all joints of Peking Opera performances is calculated. The calculation formula is as follows:

[0044] Therefore, the static similarity score Similarity can be calculated by the following formula:

[0045] Where n represents the number of joints. Multiplying by 100 in the formula can convert the static similarity score Similarity into a percentage system, making the data more intuitive.

[0046] The following is a description of the dynamic action evaluation unit processing a sample of Peking Opera performance actions.

[0047] Based on the angles of each non-terminal joint point of the Peking Opera performance action and the angles of each non-terminal joint point of the corresponding reference action, a time-series angle data sequence is constructed, and a DTW distance matrix of size 'M×N' is constructed for each pair of non-terminal joint points, where 'M' and 'N' are the time lengths of the Peking Opera performance action and the pre-stored corresponding reference action time series. Each element '(i, j)' of the matrix represents the angle difference between the angle of the joint point of the Peking Opera performance action of the i-th time length in sequence one and the angle of the joint point of the j-th time length reference action. The absolute value of the difference between the corresponding joint point angles at each moment in the action sequence is calculated and used as the element of the matrix to fill the matrix.

[0048] The constructed angle distance matrices of each joint point are summed and averaged. First, the values ​​of the elements at the same position of all matrices are accumulated and then averaged, and finally a comprehensive distance matrix S is obtained. Starting from the upper right corner of the comprehensive distance matrix, reverse back to the lower left corner of the matrix, the cumulative distances of different paths are obtained, and the path with the smallest cumulative distance is obtained as the best time alignment path. Based on the best time alignment path, the time axes of the two action sequences are normalized to align them in time. Among them, the normalized distance represents the cumulative distance of the path starting from the upper right corner of the comprehensive distance matrix and reversely tracing back to the lower left corner of the matrix after the time axes of the two action sequences are normalized. Based on the normalized distance and the above-mentioned static similarity score Similarity, the modified dynamic action similarity score can be obtained. The modified dynamic action similarity score can also be called the preliminary similarity score of Peking Opera performance actions.

[0049] The following is a description of a sample of the output unit processing Peking Opera performance movements.

[0050] Taking a specific Peking Opera performance as an example, each non-terminal joint point of the Peking Opera performance has a corresponding weight. In the output unit, each particle in the particle swarm optimization algorithm can represent a weight configuration scheme of a non-terminal joint point.

[0051] Set the number of non-terminal joints of the Peking Opera performance action to N, generate a group of N particles, where the position of particle i Represents a specific non-terminal joint weight configuration scheme. Is a set containing the weights of each parameter, expressed as = [weight 1, weight 2…weight n], where the initial value of each weight is 1, and random perturbations are added to adjust the initial value of the weight to introduce randomness. At the same time, the speed of each particle is randomly initialized .

[0052] The fitness value of each particle is calculated in the above way. For each particle, if the fitness value of the current position is lower than the fitness value of its historical optimal position, its individual optimal position is updated. , and at the same time, the global optimal position Update to the position of the particle with the lowest fitness value in the particle group. For each particle, update its speed and position according to the current speed, individual optimal position and global optimal position. Until the number of iterations reaches the threshold or the fitness value of the predicted similarity score and the similarity score is less than the threshold

[0053] The random initialization speed of particles can be determined based on the following formula:

[0054] Among them, t is the number of iterations; ω is the inertia weight; c1 and c2 are learning factors; r1 and r2 are numbers generated by random functions.

[0055] Specifically, r1 and r2 are numbers generated by a random function to add random perturbations to the velocity update process.

[0056] The parameters of the output unit are used to implement the joint weight configuration scheme of the Peking Opera performance movements. Therefore, during the iteration process, the parameters of the output unit are adjusted as the joint weight configuration scheme changes. The final parameters of the output unit correspond to the optimal joint weight configuration scheme after iteration.

[0057] In this embodiment, the static action similarity score of each non-terminal joint point of the sample of the Peking Opera performance action is obtained through the static action evaluation unit, and the static action similarity score of each non-terminal joint point is corrected through the dynamic action evaluation unit to obtain the corresponding dynamic action similarity score, which can reduce the similarity score error caused by the length of the action sequence. The importance of each non-terminal joint point is learned through the output unit, and the weight of the dynamic action similarity score of each non-terminal joint point is adjusted, so that a more accurate predicted similarity score of the Peking Opera performance action can be output.

[0058] When adding random disturbance to adjust the initial value of the weight to introduce randomness, obtaining the joint weight configuration scheme of the Peking Opera performance action, and adjusting the parameters of the output unit, strong randomness is exhibited in the initial stage, which may cause the subsequent iterations to converge slowly. To solve this problem, based on any of the above embodiments, the sample of the Peking Opera performance action and the dynamic action similarity score are sent to the output unit to obtain the predicted similarity score of the Peking Opera performance action, including: Sending the sample of the Peking Opera performance action and the dynamic action similarity score to the output unit to determine the change activity of each non-terminal joint point of the Peking Opera performance action; According to the change activity of each non-terminal joint point, the dynamic action similarity score of each non-terminal joint point is assigned corresponding weights, and the predicted similarity score of Peking Opera performance actions is obtained by adding them up.

[0059] For example, the angle change of the target non-terminal joint relative to the previous moment can be obtained at fixed intervals, and the cumulative angle change during the action time is used as the change activity of the target non-terminal joint to obtain the change activity of each non-terminal joint. The fixed time can be 0.1 seconds, 0.2 seconds, etc., and is not further limited here.

[0060] The change activity of non-terminal joints can be sorted, and each non-terminal joint can be divided into three levels according to the proportion: relatively active, intermediate level and less active. The specific proportions of relatively active, intermediate level and less active can be 0.2, 0.5 and 0.3, or 0.3, 0.5 and 0.2, which are not further limited here. Exemplarily, the weight of the relatively active joints can be set to 1.5, the weight of the intermediate level can be set to 1.0, and the weight of the less active joints can be set to 0.5, and the initial predicted similarity score of the Peking Opera performance movements can be obtained after adding them together.

[0061] Compared with the initial joint point weight configuration scheme of Peking Opera performance movements, which is that the weights of each joint point are the same, in this embodiment, based on the change activity of each non-terminal joint point of Peking Opera performance movements, the weights of each joint point of Peking Opera performance movements are preliminarily adjusted as the initial joint point weight configuration scheme, which can make the weight of each joint point more closely fit the optimal joint point weight configuration scheme, speed up the convergence of subsequent iterations, and shorten the number of iterations.

[0062] Based on any of the above embodiments, the static action evaluation unit and the dynamic action evaluation unit respectively include non-terminal joint angle data of the reference action; Inputting the motion capture data into the action evaluation model and outputting the similarity score of the Peking Opera performance action to be evaluated includes: Inputting the motion capture data into the static motion evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action to be evaluated and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static motion similarity score of the non-terminal joint points of the Peking Opera performance action to be evaluated and the pre-stored corresponding reference action; Sending the motion capture data and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The motion capture data and the dynamic action similarity score are sent to the output unit, and the similarity score of the Peking Opera performance action to be evaluated is output.

[0063] In this embodiment, the static action similarity score of each non-terminal joint point of the sample of the Peking Opera performance action is obtained through the static action evaluation unit, and the static action similarity score of each non-terminal joint point is corrected through the dynamic action evaluation unit to obtain the corresponding dynamic action similarity score. Based on the importance of each non-terminal joint point, the output unit outputs a more accurate predicted similarity score of the Peking Opera performance action on the basis of the dynamic action similarity score of each non-terminal joint point.

[0064] The following is a description of the Peking Opera performance action evaluation device based on motion capture data provided by the present invention. The Peking Opera performance action evaluation device based on motion capture data described below and the Peking Opera performance action evaluation method based on motion capture data described above can be referenced to each other.

[0065] Figure 2 is a schematic diagram of the structure of the Peking Opera performance action evaluation device based on motion capture data provided by the present invention. Figure 2 As shown, the device comprises: The motion capture data acquisition module 201 is used to acquire the motion capture data of the Peking Opera performance movements to be evaluated.

[0066] The similarity score acquisition module 202 is used to input the motion capture data into the action evaluation model and output the similarity score of the Peking Opera performance action to be evaluated; wherein the action evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions and corresponds one-to-one to the samples of the Peking Opera performance actions.

[0067] Based on any of the above embodiments, the samples of Peking Opera performance movements include samples of Peking Opera performance movements and similarity scores thereof; the movement evaluation model includes: a static movement evaluation unit, a dynamic movement evaluation unit and an output unit.

[0068] The Peking Opera performance action evaluation device based on motion capture data also includes: a model training module for: Processing the sample of the Peking Opera performance action through the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and outputting the predicted similarity score of the Peking Opera performance action; The predicted similarity score is compared with the similarity score, and the parameters of the output unit are adjusted according to the comparison result, until the number of iterations reaches a threshold or the fitness value of the predicted similarity score and the similarity score is less than a threshold.

[0069] Based on any of the above embodiments, the model training module is specifically used for: Inputting the sample of the Peking Opera performance action into the static action evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static action similarity score of the corresponding non-terminal joint points of the Peking Opera performance action and the pre-stored corresponding reference action; Sending the sample of the Peking Opera performance action and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The samples of the Peking Opera performance movements and the dynamic movement similarity scores are sent to the output unit to obtain the predicted similarity scores of the Peking Opera performance movements.

[0070] Based on any of the above embodiments, the model training module is further used for: Sending the sample of the Peking Opera performance action and the dynamic action similarity score to the output unit to determine the change activity of each non-terminal joint point of the Peking Opera performance action; According to the change activity of each non-terminal joint point, the dynamic action similarity score of each non-terminal joint point is assigned corresponding weights, and the predicted similarity score of Peking Opera performance actions is obtained by adding them up.

[0071] Based on any of the above embodiments, the model training module is further used for: The predicted similarity score of each Peking Opera performance action is compared with the similarity score of the corresponding Peking Opera performance action to obtain the difference; According to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, the average value of the difference between the predicted similarity score of all Peking Opera performance actions and the similarity score of the corresponding Peking Opera performance action is obtained as the fitness value of the predicted similarity score and the similarity score; The parameters of the output unit are adjusted according to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, until the number of iterations reaches a threshold or the average value of the difference between the predicted similarity score of the Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action is less than a threshold.

[0072] Based on any of the above embodiments, the similarity score acquisition module 202 is specifically used to: Inputting the motion capture data into the static motion evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action to be evaluated and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static motion similarity score of the non-terminal joint points of the Peking Opera performance action to be evaluated and the pre-stored corresponding reference action; Sending the motion capture data and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The motion capture data and the dynamic action similarity score are sent to the output unit, and the similarity score of the Peking Opera performance action to be evaluated is output.

[0073] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the Peking Opera performance action evaluation method based on motion capture data, the method comprising: obtaining motion capture data of the Peking Opera performance action to be evaluated; inputting the motion capture data into the action evaluation model, and outputting the similarity score of the Peking Opera performance action to be evaluated; wherein the action evaluation model is obtained after training based on the samples of the Peking Opera performance action and the corresponding similarity score, and the similarity score is predetermined according to the samples of the Peking Opera performance action and corresponds one-to-one with the samples of the Peking Opera performance action.

[0074] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0075] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the Peking Opera performance action evaluation method based on motion capture data provided by the above-mentioned methods, the method including: obtaining motion capture data of the Peking Opera performance action to be evaluated; inputting the motion capture data into a motion evaluation model, and outputting a similarity score of the Peking Opera performance action to be evaluated; wherein the motion evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions, and corresponds one-to-one to the samples of the Peking Opera performance actions.

[0076] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the Peking Opera performance action evaluation method based on motion capture data provided by the above-mentioned methods, the method comprising: obtaining motion capture data of the Peking Opera performance action to be evaluated; inputting the motion capture data into a motion evaluation model, and outputting a similarity score of the Peking Opera performance action to be evaluated; wherein the motion evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions, and corresponds one-to-one to the samples of the Peking Opera performance actions.

[0077] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0078] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating Peking Opera performance movements based on motion capture data. It is characterized in that include: Obtain motion capture data of the Peking Opera performance movements to be evaluated; Inputting the motion capture data into a motion evaluation model, and outputting a similarity score of the Peking Opera performance motion to be evaluated; The action evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores. The similarity scores are predetermined based on the samples of Peking Opera performance actions and correspond one-to-one with the samples of Peking Opera performance actions.

2. The method for evaluating Peking Opera performance movements based on motion capture data according to claim 1, It is characterized in that The action evaluation model includes: a static action evaluation unit, a dynamic action evaluation unit and an output unit; The training method of the action evaluation model includes: Processing the sample of the Peking Opera performance action through the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and outputting the predicted similarity score of the Peking Opera performance action; The predicted similarity score is compared with the similarity score, and the parameters of the output unit are adjusted according to the comparison result, until the number of iterations reaches a threshold or the fitness value of the predicted similarity score and the similarity score is less than a threshold.

3. The method for evaluating Peking Opera performance movements based on motion capture data according to claim 2, It is characterized in that The sample of the Peking Opera performance action is processed by the static action evaluation unit, the dynamic action evaluation unit and the output unit in sequence, and the predicted similarity score of the Peking Opera performance action is output, including: Inputting the sample of the Peking Opera performance action into the static action evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static action similarity score of the corresponding non-terminal joint points of the Peking Opera performance action and the pre-stored corresponding reference action; Sending the sample of the Peking Opera performance action and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The samples of the Peking Opera performance movements and the dynamic movement similarity scores are sent to the output unit to obtain the predicted similarity scores of the Peking Opera performance movements.

4. The method for evaluating Peking Opera performance movements based on motion capture data according to claim 3, It is characterized in that Sending the sample of the Peking Opera performance action and the dynamic action similarity score to the output unit to obtain the predicted similarity score of the Peking Opera performance action includes: Sending the sample of the Peking Opera performance action and the dynamic action similarity score to the output unit to determine the change activity of each non-terminal joint point of the Peking Opera performance action; According to the change activity of each non-terminal joint point, the dynamic action similarity score of each non-terminal joint point is assigned corresponding weights, and the predicted similarity score of Peking Opera performance actions is obtained by adding them up.

5. The method for evaluating Peking Opera performance movements based on motion capture data according to claim 2, It is characterized in that Comparing the predicted similarity score with the similarity score, and adjusting the parameters of the output unit according to the comparison result, includes: The predicted similarity score of each Peking Opera performance action is compared with the similarity score of the corresponding Peking Opera performance action to obtain the difference; According to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, the average value of the difference between the predicted similarity score of all Peking Opera performance actions and the similarity score of the corresponding Peking Opera performance action is obtained as the fitness value of the predicted similarity score and the similarity score; The parameters of the output unit are adjusted according to the difference between the predicted similarity score of each Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action, until the number of iterations reaches a threshold or the average value of the difference between the predicted similarity score of the Peking Opera performance action and the similarity score of the corresponding Peking Opera performance action is less than a threshold.

6. The method for evaluating Peking Opera performance movements based on motion capture data according to claim 2, It is characterized in that The static action evaluation unit and the dynamic action evaluation unit respectively include non-terminal joint angle data of the reference action; Inputting the motion capture data into the action evaluation model and outputting the similarity score of the Peking Opera performance action to be evaluated includes: Inputting the motion capture data into the static motion evaluation unit, obtaining and comparing the non-terminal joint angle data of the Peking Opera performance action to be evaluated and the non-terminal joint angle data of the pre-stored corresponding reference action, and obtaining the static motion similarity score of the non-terminal joint points of the Peking Opera performance action to be evaluated and the pre-stored corresponding reference action; Sending the motion capture data and the static action similarity score to the dynamic action evaluation unit, performing time alignment on the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action to obtain a normalized distance between the Peking Opera performance action to be evaluated and the corresponding pre-stored reference action, and correcting the static action similarity score based on the normalized distance to obtain a corresponding dynamic action similarity score; The motion capture data and the dynamic action similarity score are sent to the output unit, and the similarity score of the Peking Opera performance action to be evaluated is output.

7. A Peking Opera performance action evaluation device based on motion capture data, It is characterized in that include: A motion capture data acquisition module, used to acquire motion capture data of the Peking Opera performance movements to be evaluated; A similarity score acquisition module is used to input the motion capture data into an action evaluation model and output a similarity score of the Peking Opera performance action to be evaluated; wherein the action evaluation model is obtained after training based on samples of Peking Opera performance actions and corresponding similarity scores, and the similarity score is predetermined based on the samples of the Peking Opera performance actions and corresponds one-to-one to the samples of the Peking Opera performance actions.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the method for evaluating Peking Opera performance movements based on motion capture data as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method for evaluating Peking Opera performance movements based on motion capture data as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the method for evaluating Peking Opera performance movements based on motion capture data as claimed in any one of claims 1 to 6 is implemented.

Images