一种基于隐马尔可夫模型的人脸识别方法

A FACE RECOGNITION APPROACH BASED ON HIDDEN MARKOV MODEL

提出一种改进的基于隐马尔可夫模型的人脸识别方法。利用人脸隐马尔可夫模型的结构特征和Viterbi算法的特点,对特征观察序列进行分割,使用部分序列对所有隐马尔可夫模型递进地计算最大相似度,同时排除相似度最小的隐马尔可夫模型,减少观察序列的计算次数,提高识别效率。实验结果表明,该方法能在不降低识别率的情况下,有效提高识别速度。

An improved approach for face recognition based on improved hidden Markov model is proposed.This approach makes use of the structural feature of hidden Markov model and the characteristic of Viterbi algorithm to segment the feature observation sequence,uses part of the sequence to calculate in progressive way the maximum similarities between all the hidden Markov models,and at the same time eliminates those models having least similarities so as to reduce the calculation times of the observation sequence and to increase the recognition efficiency.Results of the experiments show that this approach can effectively improve the recognition speed under the condition of no impact on the recognition rate.