Application of FTGSVM Algorithm in Expression Recognition of Fatigue Driving

Fatigue driving is a main cause for road traffic accidents. In case an alarm can be sent to drivers one second before the potential traffic accidents, 90% similar traffic accidents can be avoided. Facial expression recognition of drivers is an important component of fatigue driving prewarning system. For the continuous video images obtained, facial features of eyes, nostril and mouth are firstly acquired via horizontal projection curve to construct the facial feature triangle. Then potential rectangular region of face tracking is generated via rigid constraints. Within the rectangular region of feature tracking, two-dimensional Gabor kernel function is chosen to construct 48 optimal filters, obtain 48 eigenvalues, and carry out training via SVM. Experiment result shows that the recognition rate of the algorithm can be as high as 94.8% for the most common neutral expression. The average recognition rate can be as high as 93.9% for the seven expressions, namely, happy, surprise, sad, angry, fear, and disgust.