Image recognition algorithm of aerobics athletes' upper limb movements based on federated learning

In order to objectively evaluate and feedback the actions of aerobics athletes and improve the level of auxiliary training and performance, an image recognition algorithm of aerobics athletes' upper limb actions based on Federated learning was proposed. The images of upper limb movements of aerobics athletes were collected, and the Gaussian function was used to denoise the images of upper limb movements of aerobics athletes. Using Zernike moments, extract the upper limb movement image features of aerobics athletes. On this basis, based on the Federated learning algorithm, the image feature model of aerobics athletes' upper limb actions is trained to realize the image recognition of aerobics athletes' upper limb actions. The experimental results show that this method has a maximum recognition accuracy of 98.4 % for upper limb movement images of aerobic athletes, with a recognition time of less than 6.1, and has good recognition performance.