Dynamic Gesture Recognition Method Based on Convolutional Neural Network

Abstract

A method for identifying dynamic gestures is studied for the problem of low recognition accuracy under the rotation angle in gesture recognition of human-computer interaction. The methodology of the design is summarized in four processes. Firstly, the inter-frame difference method is used to extract the gesture key frame. Secondly, Hue Saturation Value (HSV) color space model is used to achieve gestures positioning. Thirdly, gesture key frame and final gesture frame are stitched together. In the final process, training and testing data using convolutional neural network model. The experimental results show that the recognition accuracy of the proposed five kinds of gesture models can reach 92%. The method can better realize the recognition of the gesture under the rotation angle.