基于隐马尔科夫模型的人脸识别

Face Recognition Based on Hidden Markov Model

对于人脸识别系统来说,人脸图像的特征提取和匹配是决定人脸识别系统性能的关键所在。文中提出基于隐马尔科夫模型的人脸识别方法。首先,根据人脸的特点建立马尔科夫模型,然后对图像进行预处理,再利用采样窗对人脸图像进行采样并进行离散余弦变换,提取变换后的系数作为观察向量。最后对人脸图像进行HMM训练,训练结束后即建立了一个人的HMM。基于DCT系数的二维隐马尔科夫模型由于充分利用了人脸图像的二维统计特性,具有较高的识别率。实验结果证明此方法在准确性方面具有良好的性能。

Extracting face features from facial images and facial image matching are the most important part in the face recognition system.A face recognition approach based on Hidden Markov Model was presented in this paper.Firstly,this method establishes the Hidden Markov Model according to the features of face,then preprocessing the image,sample the face image by sampling window and process DCT（Discrete Cosine Transform） transformation,get the DCT coefficient as observation vector.Finally train the face image,after trained the image,get the HMM parameters of a people.The face recognition algorithm based on DCT coefficient makes the full use of face image 2-D statistical property,with high rate in different condition.The experimental results demonstrate that the proposed algorithm has excellent performance.