An Eye States Detection Method by Using WLBP

In this paper, we presents an efficient eye states detection method by using WLBP. WLBP is a novel feature extraction method we proposed, which combines the advantages of WLD and LBP. For an eye image, its WLBP histogram is extracted and each bin of the histogram is regarded as a feature of the eye. By using the extracted eye features and Support Vector Machines (SVM), a non-linear classifier is trained to recognize the eye state. Experimental results show that WLBP is obviously superior to WLD and LBP. Meanwhile, it´s also robust to noise and light variation. The experiments under inside-car environment also demonstrate the effectiveness and robustness of our method.