Abstract: In view of problem of low recognition accuracy caused by sudden change of illumination, change of face view angle and external obstacle, a face recognition method based on SIFT feature in hidden Markov model(HMM) was proposed. The method takes extracted SIFT feature as training sample of HMM, and gets accurate matching library which classifies SIFT feature of face and non-face, so as to realize recognition of face image. The experimental results show that the method has quick recognition speed and high accuracy with average recognition accuracy rate of 97.14%.