一种基于嵌入式隐马尔可夫模型的人脸识别方法  

An approach for face recognition based on embedded hidden Markov model

作　　者：沈杰[1] 王正群[1] 邹军[1] 侯艳平[1] 

机构地区：[1]扬州大学信息工程学院,江苏扬州225009

出　　处：《扬州大学学报（自然科学版）》2008年第1期59-62,共4页Journal of Yangzhou University：Natural Science Edition

基　　金：江苏省高校自然科学基金资助项目(05KJB520102;07KJB520133);扬州大学自然科学基金资助项目(KK0413160)

摘　　要：提出一种基于嵌入式隐马尔可夫模型(embedded hidden Markov model,简称EHMM)的人脸图像识别方法,主要包括:1由于奇异值向量具有稳定性、转置不变性等特点,故对归一化的人脸图像,可采用奇异值分解抽取人脸图像特征作为观察序列;2在人脸识别中应用嵌入式隐马尔可夫模型,采用多高斯概率密度函数训练、建立EHMM,再利用建好的EHMM进行识别.实验结果显示,所提方法减少了数据计算量,运行速度快,并提高了识别率,完全满足人脸识别系统的实时性要求.A human face recognition approach based on the embedded hidden Markov model is proposed and its essential contents list as follows： （1） With the stability and the transposition invariability of the singular value vector, singular value decomposition is used to extract the character of the face image as the observed value array for the normalized face image; （2） In the face automatic recognition system, an EHMM （embedded hidden Markov model） method is used to recognize face, the multiplicate Gaussian pdf is adopted to build up the EHMM, and then the model could be used to identify face. The experimental results show that the method proposed reduces the computing time, runs faster, and the recognition rate is increased, so it can satisfy the needs of real-time face recognition.

关 键 词：模式识别 嵌入式隐马尔可夫模型 奇异值分解 人脸识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]