针对人脸识别的鲁棒性问题，鉴于HMM 具有良好的时间序列建模能力和SVM 在有限样本的分类方面具有优良性能，采用一种基于HMM-SVM 融合模型的鲁棒人脸识别算法。首先将归一化人脸图像用采样窗从上到下进行采样，采用DCT 和SVD 提取各个采样窗图像的特征参数并串接成观察向量，然后由每个人的训练图像的观察向量训练得到每个人HMM 模型，将测试图像的观察向量采用Viterbi 算法求出对应于每个人HMM 模型的输出概率，最后将输出概率送入支持向量机进行分类训练及识别测试，得到人脸识别结果。在ORL 库和Yale 库的实验表明该算法的识别率高于传统的单一HMM 方法和SVM 方法，鲁棒性有一定的提高。

To deal with the robust face recognition problem, a mixed model based on Hidden Markov Model (HMM) and Support Vector Machine (SVM) was proposed for HMM has good ability for time sequence modeling and SVM has excellent ability for classifying. Firstly, a sequence of overlapping sub-images was extracted from face image by using Discrete Cosine Transform (DCT) and Singular Value Decomposition (SVD). Then, the sequence which was extracted from training images was modeled by using HMM, and the output probability of each HMM for the training sequence had been considered as the input vector of SVM for its training. Finally, the output probability of each HMM for the testing sequences had been considered as the input vector of SVM for its testing. Experimental results on ORL and Yale face database demonstrate that the effectiveness and robustness of the proposed algorithm are better than traditional signal HMM and SVM algorithm.