Research On Human Body Gesture Recognition Algorithm Based On Convolutional Neural Network

With the development of artificial intelligence,researchers have made great progress in the research of human body gesture recognition in recent years.On the basis of summarizing the existing research status of human pose recognition,the paper uses skeleton information for feature extraction instead of image processing for feature extraction.The traditional skeletal feature modeling uses the skeletal angle,angular acceleration,joint point velocity,acceleration rate,kinetic energy,potential energy and other dynamic characteristics as the modeling parameters.This makes the gesture model involve many parameters,and the disturbance of the parameters has a great influence on the recognition result,so that the robustness of the recognition is low.A human body identity and attitude modeling method based on joint coordinates of three-dimensional coordinates are proposed based on the traditional modeling method.Together with a human identity and gesture recognition algorithm with high recognition rate are proposed.The method uses a depth sensor to obtain a human skeleton map,and performs feature extraction based on joint point coordinate information.On the basis of feature extraction,a human identity model based on the length of the joint point of interest,a static attitude model based on distance method and a motion attitude model based on coordinate method are established.The stability analysis of the model is carried out through experiments.The model has the characteristics of small amount of data and high robustness.In order to solve the model accurately,the paper proposes an identification algorithm based on BP neural network and convolutional neural network.The neural network structure,corresponding activation function and main parameters were determined by theoretical analysis and experimental verification.The proposed algorithm has the characteristics of less training samples and faster training rate.After the neural network model is trained,the proposed algorithm can accurately determine the human identity and human posture.The actual experiment shows that the recognition success rate of the human body's static posture reaches 96%,and the recognition success rate of human body motion posture and human body identity reaches 98%.Therefore,the proposed model and algorithm are feasible and effective,and the proposed identification scheme is characterized by strong operability and easy implementation.The paper developed a human body gesture recognition system with VSdevelopment tools.