中文摘要：

提出了一种优化嵌入式隐Markov模型状态数的搜索算法，并且由此算法得到改进的嵌入式隐Markov模型结构。实验证明，这种结构既能提高人脸的识别率，最高可达到100％，又比传统的结构减少30％左右的训练时间和识别时间。

英文摘要：

An improved embedded hidden Markov model structure for face recognition is presented. The model not only improves the recognition rate, but also reduces the train and recognition time costs dramatically. Experimental results show that the highest recognition rate of 100% is achieved. The most exciting thing is that the structure can shrink 30% time costs than traditional structure both on train and recognition stage.