This paper constructs a motion correction and training effect improvement system for sports athletes driven by digital technology. High-speed video acquisition, inertial sensing, plantar pressure and heart rate load data are used to synchronize modeling, and a multi-source representation for special training is formed. At the recognition end, the system introduced the key point timing correlation and deviation quantification mechanism to distinguish the joint trajectory, the center of gravity transfer, the force rhythm and the action stability. At the control end, the hierarchical feedback of training phase, load status and completion quality output was combined to realize the linkage update of correction tips, rhythm correction and intensity adjustment. Experimental results based on 5504 training samples and 200 athletes show that the deviation detection rate of the system in six types of training movements is above 95.9%, the correction accuracy is between 95.4%-97.0%, and the average response delay is maintained at about 78 ms. After 6 weeks of continuous training, the improvement rate of movement completion in the system assisted training group reached 11.7%-13.4%, and the average score of special training increased from 82.4 to 91.3. In the continuous training scenario, the success rate of the system is kept between 94.8% and 97.2%, which reflects good field adaptability and real-time operation performance.
Povzetek: In the experiment of 5504 groups of training samples and 200 athletes, the deviation detection rate of more than 95.9% was achieved, the average response delay was maintained at about 78 ms, and the movement completion improvement rate reached 11.7%-13.4% after 6 weeks of training. It can provide stable technical support for improving the training effect of athletes.
