The Research On The Perception Of Driver’s State Based On Compressive Sensing

It is well known that the driver is the main cause of traffic accident. Thus, the real-timemonitoring of the driver emotion or fatigue is of vital practical significance. This paper hasmainly researched the perception of driver’s state as follows:Based on the study of compressive sensing, this paper presented an algorithm for driver’smulti-angle face conversion based on position priori and sparse representation, an algorithmfor driver’s facial expression recognition based on sparse representation and an algorithm fordriver’s eye state recognition based on sparse representation. All the above algorithms aresparse representation of the samples to be tested, using a minimum L-1norm solving thesparse representation coefficients.In the part of the algorithm for driver’s multi-angle face conversion which based on positionpriori and sparse representation, the face angle is transformed by using the sparse coefficientsand multi-angle face database, and the experimental results show that the algorithmframework is relatively effective.In the part of algorithm for driver’s facial expression recognition based on sparserepresentation, the sparse representation classifier is created by using the class information ofsparse coefficients, and thus the facial expression image of drivers to be tested can beclassified. During the identification process, due to the huge amount of data, highcomputational complexity, this article applied two popular dimension reduction methods-downsampling and principal component analysis–to reduce the amount of data, and put thedimension-reduced feature image into the algorithm framework. The person-dependent andperson-independent facial expression recognition experiments which based on standard facedatabase have obtained reasonable results. Moreover, according to the actual situation of thedrivers, the normal, eye closing, yawning and smiling facial images is collected and tested,and the accuracy of recognition are relatively high.In the part of algorithm for driver’s eye state recognition that based on sparserepresentation, the eye images of drivers to be tested are classified and recognized by sparserepresented classifier. What’s more, the eye state simulations of eye-opening and eye-closingacquired ideal results.