Abstract: Driving in case of drowse is a most important factor of traffic accidents.Under the circumstances of the natural driving，drowsing driving and fatigued driving of a driver，the opening degrees of the eyes are different to some extent.This paper proposes a recognition algorithm to the eyes’ states based on ICA and Hidden Markov Model（HMM）.The original image is used to get edge grayscale image and edge grayscale enhanced image.Secondly，it is applied to extraction of eye state features using ICA algorithm，in this process it adopts FastICA algorithm in order to increase the speed of feature extraction.Finally，it is applied to recognizing eye state using HMM algorithm.Experimental results show that the system increases the whole effect and accuracy of eye state recognition，and prove that the algorithm is efficient and feasible.