基于傅立叶描述子-BP神经网络的手势识别方法

Hand Gesture Recognition Based on Fourier Descriptor-BP Neural Network

针对手势识别中人手是复杂变形体,手部特征描述容易受到环境因素影响的特点,提出了一种基于傅立叶描述子-BP神经网络的手势识别方法.首先根据YCbCr和Nrg肤色模型的互补性以及背景模型有效去除复杂背景中的类肤色的特点,利用多特征相融合的手势分割方法提取手势区域;然后结合傅立叶描述子具有较好的轮廓描述能力和BP神经网络较强的自学习能力,利用傅立叶描述子-BP神经网络方法对手势进行识别.实验结果表明此方法具有较好的鲁棒性和较高的识别率.

Hand is a highly variable organ and hand features are easily affected by environmental factors. Considering the characteristics of hand gesture, a novel hand gesture recognition algorithm based on Fourier descriptor and BP neural network is presented. According to the complementarity of the YCbCr and Nrg color models and the background model that is available to exclude the similar skin colors from the complicated background, the hand gesture region is extracted by the segmentation of hand gestures through multi-feature integration. Then, the advantages of the Fourier descriptor which can describe well the profiles and of the BP neural network which is strong at self-learning are taken and combined together to recognize hand gestures. The experimental results showed that the method proposed has better robustness and higher recognition rate.