Abstract: This paper presents an approach for Hidden Markov Model(HMM) training based on Particle Swarm Optimization(PSO). HMM parameter adjustment rules are induced by using PSO and Quantum-behaved Particle Swarm Optimization(QPSO) separately. Discriminative information contained in the training data is used to improve the performance on HMM effectively and the method is used in facial expression recognition. Facial expression feature vectors are extracted by using Discrete Cosine Transform(DCT). Experimental results show that the new method provides satisfactory recognition performance and is powerful for HMM parameter estimation.

Key words: Hidden Markov Model(HMM), Particle Swarm Optimization(PSO), Discrete Cosine Transform(DCT), facial expression recognition

摘要： 提出基于微粒群优化算法(PSO)的隐马尔科夫模型(HMM)训练算法，分别用PSO和量子微粒群优化算法进行HMM的参数估计，以提高HMM的性能。将改进的HMM算法应用于人脸表情识别，采用离散余弦变换提取表情特征向量。实验结果表明，该算法能有效提高表情识别率，解决HMM的参数估计问题。

关键词: 隐马尔科夫模型, 微粒群优化算法, 离散余弦变换, 表情识别

CLC Number: 

• TP301.6