A Sports Training System Based on Kinect Intelligent Wearable Device

Combined with the research of somatosensory technology and human motion recognition, this paper proposes an intelligent motion training system based on somatosensory recognition. Firstly, this paper filters the bone data collected by Kinect through jitter elimination and double exponential smoothing, and then extracts the feature vector of bone data by using space vector method to build a finite state machine model. The DTW algorithm combined with the weighted Euclidean distance is used to match the motion template to achieve the accurate and low delay motion recognition target. This paper first analyzes the design requirements of the system, designs the hardware architecture and software architecture of the system according to the requirements, and makes a detailed description of the specific implementation of motion recognition in the software. The experimental results show that after 198 tests, the motion recognition rate of the pull-up is 100
^%
, and the accuracy rate is 99%. After 200 tests, the motion recognition rate of the squat is 99%, and the accuracy rate is 97.5%. After 594 tests, the motion recognition rate of the standing long jump is 97.8
^%
, and the accuracy rate is 87.9
^%
. It is proved that the recognition rate of the system for motion is very high and the motion recognition function is very reliable, but the accuracy rate can reach the level of consumer grade products except for standing long jump. Result finding that although the system still has some drawbacks in the uninterrupted long jump performance, the success of the overall bow motion recognition algorithm indicates good motion recognition performance without rapid movement, can accurately identify the position and motion-decomposing motions.