S.S. Tao, K.B. Chen

Pages: 151-162

Abstract
Accurate identification of fatigue driving behavior is crucial for improving road safety and preventing accidents. To effectively caution against fatigue driving, this paper proposes an intelligent recognition method for fatigue driving behavior based on wavelet energy entropy. Firstly, the anti-sharpening enhancement algorithm is employed to boost the image's contrast, details, and edges. Secondly, a multi-scale decomposition of the image is carried out to calculate the energy entropy of the wavelet coefficients at each scale, quantitatively characterizing facial feature parameters under a state of fatigue driving. Then, by focusing on blinking, yawning, and head movements, labor features are extracted based on the quantitative description of facial features. Finally, fatigue driving behavior is identified through the search for feature points and fitting results. The experimental results show that the method can effectively control the misidentification rate below 4%, with a recognition time of 690ms and a recall rate of over 97.04%, fully verify the effectiveness of the new intelligent recognition method.
Keywords: fatigue driving behavior; facial images; image enhancement; wavelet energy entropy; characteristic parameters; intelligent recognition