An eye state identification method based on the Embedded Hidden Markov Model
Huabiao Qin, Jun Liu, Tianyi Hong · 2012
This paper focuses on improving the accuracy and the speed of eye state identification, a novel method based on EHMM (Embedded Hidden Markov Model) was proposed. We extract the 2D-DCT feature of each eye image, use the low-frequency coefficients of the DCT to generate observation vector, then train the model according to the EHMM training algorithm and get classifiers. Experiment results show that when the sampling window to take 12×12, and the number of Gaussian Mixture Models to take 3, we achieve a satisfactory result. Comparing with other methods, the method presented in this paper is not sensitive to deflection angles of face and illumination. The recognition speed can be up to 20 frames/ sec so that it can be used in real system.