CN-121330567-B - Joint collaborative quantitative analysis method and device and electronic equipment

CN121330567BCN 121330567 BCN121330567 BCN 121330567BCN-121330567-B

Abstract

The invention provides a joint collaborative quantization analysis method, a device and electronic equipment, and relates to the technical field of artificial intelligence, wherein the method comprises the steps of analyzing a video to be analyzed based on BlazePose algorithm to obtain coordinate data; the method comprises the steps of determining a kinematic data matrix based on coordinate data, inputting the kinematic data matrix into a joint cooperative attention model to obtain a joint cooperative variation score, determining balance, stability, rhythmicity, reconfigurability evaluation result and maximum joint movement angle based on the kinematic data matrix, wherein the joint cooperative quantitative analysis result of a user comprises the joint cooperative variation score, the balance evaluation result, the stability evaluation result, the rhythmicity evaluation result, the reconfigurability evaluation result and the maximum joint movement angle. By the method, equipment cost and operation threshold can be reduced, application scenes of the method are enlarged, the problem of limited evaluation dimension caused by single index or score is avoided, and accuracy of joint collaborative quantitative analysis results is effectively improved.

Inventors

WANG CHEN
PENG LIANG
CHEN JINGYAO
HOU ZENGGUANG

Assignees

中国科学院自动化研究所

Dates

Publication Date: 20260508
Application Date: 20250820

Claims (8)

1. A joint cooperative quantization method, comprising: Analyzing a video to be analyzed based on BlazePose algorithm to obtain coordinate data of joint mark points, wherein the video to be analyzed is a video for a user to execute motion evaluation action; determining a kinematic data matrix for the user based on the coordinate data; Inputting the kinematic data matrix into a joint cooperative attention model to obtain joint cooperative variation degree scores output by the joint cooperative attention model; Determining a balance evaluation result, a stability evaluation result, a rhythmicity evaluation result, a reconfigurability evaluation result and a maximum joint movement angle based on the kinematic data matrix; Wherein the joint cooperative quantitative analysis result of the user comprises the joint cooperative variation score, the balance evaluation result, the stability evaluation result, the rhythmicity evaluation result, the reconfigurability evaluation result and the maximum joint movement angle; the joint collaborative attention model comprises a serial attention module, a graph convolution network pre-classification module and a cyclic mask module; the serial attention module is used for carrying out characteristic serial connection on the kinematic data matrix to generate serial characteristics carrying attention weights; the graph convolution network pre-classification module is used for generating a user pre-classification result based on the series characteristics; the cyclic mask module is used for generating the joint cooperative variation score based on the serial characteristics and the user pre-classification result; The reconfigurability evaluation result is determined by the following steps: Based on the kinematic data matrix, a space-time vector matrix is obtained through nonnegative matrix decomposition; generating a reconstructed kinematic data matrix based on the space-time vector matrix; Determining a percentage of reconstruction of the user based on the reconstructed kinematic data matrix and the kinematic data matrix; and comparing and analyzing the reconstruction percentage with the reconstruction percentage of healthy people, and determining the reconstruction evaluation result.
2. The joint cooperative quantization method of claim 1, wherein the determining a kinematic data matrix of the user based on the coordinate data comprises: Analyzing by a kinematic analysis algorithm based on the coordinate data to generate time sequence track data and calculating time sequence angle data of the joint mark points; And based on a second-order B-spline interpolation algorithm, carrying out standardization processing on the time sequence track data and the time sequence angle data to generate the kinematic data matrix.
3. The joint collaborative quantitative analysis method of claim 1, wherein the joint marker points include a left hip joint and a right hip joint; The balance evaluation result is determined by the following steps: determining a first dynamic track of the included angle between the left hip joint and the horizontal direction shaft based on the kinematic data matrix; determining a second dynamic track of the included angle between the right hip joint and the horizontal direction shaft based on the kinematic data matrix; And determining the balance evaluation result based on the first dynamic track and the second dynamic track.
4. The joint collaborative quantitative analysis method of claim 1, wherein the joint marker points include a left hip joint and a right hip joint; the stability evaluation result is determined by the following steps: determining a third dynamic track of the gravity center of the user in a sagittal plane based on the kinematic data matrix, wherein the gravity center of the user is the midpoint of a connecting line of the left hip joint and the right hip joint; Determining a gravity center track deviation standard deviation of the third dynamic track; and determining the stability evaluation result based on the gravity center track deviation standard deviation.
5. The joint cooperative quantification analysis method of claim 1, wherein the rhythmicity assessment result is determined by: Determining a left foot dynamic track and a right foot dynamic track based on the kinematic data matrix; determining a gait cycle of the user based on the left foot dynamic trajectory and the right foot dynamic trajectory; based on a dynamic time warping algorithm, aligning the left foot dynamic track and the right foot dynamic track, and determining a variance value of the gait cycle; And determining the rhythmicity assessment result based on the variance value.
6. The joint collaborative quantization method according to claim 1, wherein the joint marker points include a left hip joint, a right hip joint, a left knee joint, and a right knee joint; the maximum articulation angle is determined by: Determining a first space included angle between a first joint connecting line and a vertical direction axis based on the kinematic data matrix, wherein the first joint connecting line is a connecting line between the left hip joint and the left knee joint; determining a second space included angle between a second joint connecting line and the vertical direction axis based on the kinematic data matrix, wherein the second joint connecting line is a connecting line between the right hip joint and the right knee joint; and screening the first space included angle and the second space included angle to determine the maximum joint movement angle.
7. A joint cooperative quantization analysis device, comprising: The system comprises a coordinate data determining module, a coordinate data analyzing module and a data analyzing module, wherein the coordinate data determining module is used for analyzing a video to be analyzed based on BlazePose algorithm to obtain coordinate data of joint mark points; a kinematic data determination module for determining a kinematic data matrix of the user based on the coordinate data; The cooperative variation score determining module is used for inputting the kinematic data matrix into a joint cooperative attention model to obtain a joint cooperative variation score output by the joint cooperative attention model; the index evaluation module is used for determining a balance evaluation result, a stability evaluation result, a rhythmicity evaluation result, a reconfigurability evaluation result and a maximum joint movement angle based on the kinematic data matrix; Wherein the joint cooperative quantitative analysis result of the user comprises the joint cooperative variation score, the balance evaluation result, the stability evaluation result, the rhythmicity evaluation result, the reconfigurability evaluation result and the maximum joint movement angle; the joint collaborative attention model comprises a serial attention module, a graph convolution network pre-classification module and a cyclic mask module; the serial attention module is used for carrying out characteristic serial connection on the kinematic data matrix to generate serial characteristics carrying attention weights; the graph convolution network pre-classification module is used for generating a user pre-classification result based on the series characteristics; the cyclic mask module is used for generating the joint cooperative variation score based on the serial characteristics and the user pre-classification result; the reconfigurability evaluation result is determined by obtaining a space-time vector matrix through nonnegative matrix factorization based on the kinematic data matrix; generating a reconstructed kinematic data matrix based on the space-time vector matrix; Determining a percentage of reconstruction of the user based on the reconstructed kinematic data matrix and the kinematic data matrix; and comparing and analyzing the reconstruction percentage with the reconstruction percentage of healthy people, and determining the reconstruction evaluation result.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements the joint cooperative quantitative analysis method of any of claims 1 to 6 when the computer program is executed.

Description

Joint collaborative quantitative analysis method and device and electronic equipment Technical Field The invention relates to the technical field of artificial intelligence, in particular to a joint collaborative quantitative analysis method, a device and electronic equipment. Background The problem of dyskinesia of the human body is increasingly prominent due to the frequency of various diseases and sports injuries, which directly leads to abnormal changes in the movement pattern of the human body. The scientific community has demonstrated that the Central Nervous System (CNS) of the human body can regulate the motor pattern of the human body through a specific synergistic mechanism, which has remarkable plasticity characteristics, which means that the recovery of the nerve function and the motor ability of the human body can be effectively promoted through timely and accurate rehabilitation intervention. In this context, it is particularly important to build a comprehensive and quantifiable set of motion synergy assessment systems. Currently, related research has focused mainly on the intelligent retrofitting of traditional rehabilitation ratings scales (e.g., fugl-Meyer ratings, range of motion scales, berg balance scales, etc.). In some related art, motion data of a human body can be collected and automatically evaluated through a virtual reality system and a wearable device, and qualitative analysis research is performed based on the human body motion data and physiological data. In the aspect of joint collaborative quantization analysis, the mainstream analysis method is to process and analyze human motion data through feature decomposition technologies such as a Principal Component Analysis (PCA) method, a non-Negative Matrix Factorization (NMF) method and the like so as to obtain a joint collaborative quantization analysis result. However, the analysis methods have obvious defects in processing nonlinear data and related task variables, are easy to generate an overfitting problem, and are difficult to capture the dependency relationship between space-time characteristics and long-term motion at the same time, so that the accuracy of joint collaborative quantitative analysis results is insufficient. In addition, the existing motion function analysis system generally adopts wearing or patch type optical motion capture equipment and an inertial sensing unit to collect human motion data, and then combines a data-driven machine learning algorithm to identify an abnormal motion mode of a human body. In summary, the existing joint collaborative quantitative analysis method has inaccurate analysis results, high cost and limited application scenes. Disclosure of Invention The invention provides a joint collaborative quantization analysis method, a device and electronic equipment, which are used for solving the defects of inaccurate analysis result, high cost and limited application scene of the existing joint collaborative quantization analysis method. The invention provides a joint collaborative quantization analysis method, which comprises the steps of analyzing a video to be analyzed based on BlazePose algorithm to obtain coordinate data of joint mark points, determining a kinematic data matrix of a user based on the coordinate data, inputting the kinematic data matrix into a joint collaborative attention model to obtain joint collaborative variation scores output by the joint collaborative attention model, and determining balance evaluation results, stability evaluation results, rhythmicity evaluation results, reconfiguration evaluation results and maximum joint movement angles based on the kinematic data matrix, wherein the joint collaborative quantization analysis results of the user comprise joint collaborative variation scores, balance evaluation results, stability evaluation results, rhythmicity evaluation results, reconfiguration evaluation results and maximum joint movement angles. The joint collaborative quantization analysis method provided by the invention is used for determining a kinematic data matrix of a user based on coordinate data, and comprises the steps of analyzing the coordinate data through a kinematic analysis algorithm to generate time sequence track data, calculating time sequence angle data of joint mark points, and carrying out standardized processing on the time sequence track data and the time sequence angle data based on a second-order B-spline interpolation algorithm to generate the kinematic data matrix. The joint collaborative quantization analysis method comprises a serial attention module, a graph rolling network pre-classification module and a cyclic mask module, wherein the serial attention module is used for carrying out feature serial connection on a kinematic data matrix to generate serial features carrying attention weights, the graph rolling network pre-classification module is used for generating a user pre-classification result based on the serial features, and