Improving the Effect of Exercise through Comparison of Real and Imitated Motion Images Produced by Artificial Intelligence

We present an artificial intelligence application system to improve the skills of excise and human health. Captured video images of human exercise are processed so that the images imitates the trainer’s motion. We call such images "imitated motion images". We employed imitated motion images of a person to make it easier to imitate trainer’s exercises skills by the person. As an experiment for the evaluation of imitated motion images, we performed experiments with our collaborators. Then, we asked the collaborators for their opinions about real and imitated motion images. The results suggest that the imitated motion images are more effective. Because manual processing to produce imitated motion images is too much work, the existing technology, Pose2Pose, is used. When we tested multiple motion images of multiple person, we confirmed that about half of the images could be processed automatically, and it was also confirmed that the effect on the improvement of skills was achieved.