Typical Action Recognition Method Of Classical Cross-Country Skiing Using Inertial Sensors

Cross-country skiing is a traditional sport in the Winter Olympics,and China’s crosscountry skiing started relatively late with a low level of scientific training.In cross-country skiing training,coaches mainly rely on their experience to guide athletes’ technical movements and tactical strategies,lacking scientific quantified guidance,which leads to low efficiency.This paper proposes a typical technical movement recognition system based on inertial sensors to achieve scientific quantification of technical movements,provide data support for improving athletes’ technical movements and optimizing tactical strategies,and thereby improving the scientific training level of coaches.To meet the practical needs of coaches for quantified monitoring of technical movements in cross-country skiing training,this paper designs a traditional cross-country skiing typical technical movement recognition system based on inertial sensors.The system uses inertial sensors combined with a motion interval segmentation algorithm based on single-axis angular velocity and ant colony clustering algorithm to recognize typical technical movements in traditional cross-country skiing.GPS data is also utilized to obtain relevant parameters of typical movements.Specifically,the paper designs the overall architecture of the recognition system,including the perception layer,preprocessing layer,and recognition layer.The perception layer realizes the data flow from multiple inertial units to the mobile receiver.The preprocessing layer analyzes skiing technical movement patterns based on IMU data,designs a motion interval segmentation algorithm based on movement windows according to rules,calculates time-frequency characteristics,and verifies the segmentation algorithm using SVM algorithm.To address the uneconomical and unrealistic use of supervised learning algorithms to label data in multi-level and multi-batch athlete movement recognition,the paper designs an ant colony clustering algorithm to recognize three typical technical movements,and validates the recognition rate compared with the K-means clustering algorithm.Based on the recognition results and GPS data,movement-related parameters such as average skiing speed,skiing frequency,and maximum swing angle can be calculated.The cross-country skiing typical technical movement recognition system has the characteristics of a simple and efficient system structure.The hardware scheme of the motion recognition system uses three IMUs,which can accurately recognize athlete movements with the least wearable sensors without affecting their normal training.The application of the ant colony clustering algorithm in the recognition of multi-level and large-batch cross-country skiing movement data does not require pre-labeling of the data,and is more efficient than traditional supervised learning algorithms,thus increasing the system’s capability for large scale practical applications.By analyzing the recognition results,differences in skiing strategies and technical movements among athletes at different levels and on different terrains can be identified,which can help athletes improve their skiing strategies and technical movements,thereby improving their athletic performance and competition results.