针对现有改进的Camshift手势跟踪算法没有考虑光照变化影响下的鲁棒性, 进而降低了动态手势的识别率, 提出一种基于深度预分割结合Camshift跟踪算法的动态手势识别法。通过在Camshift手势跟踪的基础上引入深度信息, 对手势搜索区域进行深度预分割, 改进手势目标匹配概率, 去除非手势肤色区域及光照变化的影响, 最后用隐马尔可夫模型(HMM)进行识别。实验结果表明, 提出的方法在光照变化及肤色干扰的环境下有很好的鲁棒性, 数字0~9的平均识别率可达97.7%。

The existing hand gesture tracking algorithm based on modified Camshift takes no consideration of the robustness under changing illumination, thereby the recognition rate will be reduced. So in this paper, proposed is a dynamic hand gesture recognition method based on the combination of depth pre-segmentation with Camshift tracking algorithm. Firstly, the depth information was introduced into the Camshift geature tracking algorithm, then a depth pre-segmentation was given to the search region of the hand gesture, and the matching probability of target hand gesture was modified, so as to remove the influence of non-hand gesture skin area and illumination changes. Lastly, the hidden Markov model (HMM) was used for hand gesture recognition. Experimental results show that the proposed algorithm has a good robustness on illumination change and skin color interference, and the average recognition rate of number 0~9 reaches up to 97.7%.