Application of Low Resolution Face Recognition Based on Infrared Spectral Imaging in Motion Training Action Analysis

Abstract

Abstract As a conscious activity, the importance of sports in social life is self-evident. However, how to use modern scientific and technological means to improve sports fun and athletes' competitive ability has become the current research hotspot. The current sports detection system has a high restriction on Athletes' movements. Therefore, it is an inevitable direction of future development to integrate intelligent technology into sports movement detection, and face recognition technology has become the best choice with its unique technical advantages. This paper constructs an application system that combines the low-resolution face recognition technology of the Internet of things with sports training action analysis. The system determines the design idea of the system framework by optimizing the face recognition model and the embedded algorithm, and the Softmax loss function is applied to the system to optimize the recognition accuracy and credibility. In addition, the system determines the parameter response value by analyzing the elbow and knee joints in the video. Finally, according to the test of system reliability, it is found that the system with GPU as the processor has the lowest resource occupancy and the highest reliability, which indicates that the system not only avoids the problem of low accuracy of traditional identification equipment, but also makes the action detection result more reliable.