摘要
Abstract
In light of the problem of the high requirements of environment background in the gesture recognition based on vision, It presents a method of depth information in gesture extraction and recognition. Firstly, this method use the deep information provided by Kinect’s deep camera and HSV color information to segmented the palm region, and obtain a two-dimensional binary image of the palm region through mapping; Then, we can use morphological operations in the region of two-dimensional palm to processing with gesture images; And then using Hu moments to extract the features of the gesture. Finaly, using BP(Back Propagation)neural network for the gesture recognition of digital 1-10. Experimental results demonstrate that the proposed method can be adapted to the complex background, and the robustness and stability are better than the common method of gesture recognition.关键词
Kinect/手势识别/Hu矩/BP神经网络Key words
Kinect/hand gesture recognition/Hu moment/BP neural network分类
信息技术与安全科学
