A motion pattern recognition method and system based on motion training data

Abstract

The present invention provides a motion pattern recognition method and system based on motion training data, comprising the steps of: obtaining three-dimensional position information of the athlete; calculating the velocity and acceleration information of the athlete according to the obtained three-dimensional coordinate information; collecting training sample data; The speed and acceleration time series in the training sample data are preprocessed; the processed training sample data are trained using three types of recognition algorithms: binary decision tree support vector machine decision tree algorithm, BP neural network algorithm, and RBF neural network recognition algorithm. ; Use the above-mentioned three types of well-trained recognition algorithms to recognize the movement patterns of athletes, and obtain the final pattern recognition results through weighted fusion methods. The present invention has a relatively simple and convenient system structure in terms of information collection, can realize motion pattern recognition in complex and changeable training and competition scenes, and improves the accuracy and robustness of motion pattern recognition.

Description

technical field [0001] The invention belongs to the field of motion data analysis, relates to motion data collection and motion pattern recognition, in particular to a motion pattern recognition method and system based on motion training data. Background technique [0002] With the rapid development of computer software and hardware technology and network, it has become a trend to use cameras, sensors and wireless sensor networks to collect sports position information and technical and tactical data at the competition site to improve athletes' competitive level. Sports information collection and Analytical devices play an important role in this. The kinematics and dynamics data of athletes can be collected through motion tracking and positioning devices, and these data can be processed through data mining technology to obtain motion pattern characteristics, which can be used for precise sports technical and tactical analysis. [0003] Accurate sports training monitoring is ...

AI Technical Summary

Problems solved by technology

Although the above researchers have achieved a high pattern recognition rate on simple scenes, these algorithms cannot support more complex competitions and training applications with rapid changes in motion patterns

[0007] To sum up, 1. The existing motion positioning equipment is mainly divided into two types based on optics and GPS+inertial navigation. Due to its own inherent defects, the former has low motion positioning accuracy when there is occlusion in a complex motion environment, and its own system The cost is expensive, and the requirements for computer hardware are extremely high. The latter is mainly used for positioning in outdoor sports, and the positioning accuracy is not high

2. Existing motion pattern recognition algorithms mostly focus on threshold discrimination, which cannot adapt to complex motion environments for accurate recognition, and cannot better support subsequent technical and tactical analysis and motion load calculations

3. At present, there are existing indoor positioning methods based on RFID and WIFI, etc., or the positioning accuracy is not high, or the label is too large to be carried, and only two-dimensional spatial positioning can be realized, and three-dimensional spatial positioning is not supported.

Images