Human body gesture recognition method based on deep learning

Human pose recognition technology has always been an important research content in the field of computer vision and pattern recognition. So far, many pose recognition methods and theories have been proposed. However, in the process of human pose recognition, misjudgment still occurs due to the interference within the class, similar movements and other external factors. In this paper, a DBLSTM neural network model is constructed based on the methods of deep learning and data fusion for the similar motion confusion problem in the human pose recognition process. The wearable motion capture device is used to take the kinematics data of the key nodes of the human body and fuse the data with the human skeleton data extracted from the video image by Openpose. The experimental results show that the average recognition rate of similar actions of the human body is above 92%. The effectiveness of the proposed method is verified.