根據隱馬爾可夫模型(HMM)的基本理論和算法設計了一個人臉表情識別系統。該系統由兩層HMM組成：低層由六個HMM組成，分別對應六種特定表情。人臉表情特征向量進入系統后，經過低層HMM初步識別，其結果組成高層HMM的觀察向量，經過高層HMM解碼，確認出表情，從而提高了系統的識別率，增強了系統的健壯性。

Hidden Markov Model(HMM) is a widely used statistical model.This paper deal with the design of a recognition system for facial expression in the light of the principal theory and algorithm of the HMM.The system consists of two levels of HMM,with the low level composed of six HMM,corresponding to six expression-specific.The vectors of the facial expression features,after put into the system,is primarily identified through the low HMMsand result in the observation vectors of the high level of the HMM.Through deciphering of the HMM,the expression are identified,thus enhancing the recognition rate and strengthening the system to a higher level.