Recent times have seen an increased interest in technologies of driver assistance. Understanding the driver’s current status is crucial for the implementation of Advanced Driver Assistance System (ADAS) and Driver Status Monitoring (DSM). Emotional factors such as anger have been long attributed to aggressive driving behaviours and increased likelihood of road accidents. Therefore, being able to accurately detect the affective states of the vehicle occupant will be critical for enhanced safety and comfort.
In this paper, we present a methodology for the evaluation of the emotional states of vehicle drivers. The proposed approach performs an assessment of the emotional states by using combination of biologically inspired visual information processing and neural networks coupled with feedback mechanisms. The system consists of the following stages: (1) biologically inspired image pre-processing; (2) facial feature extraction; (3) multilayer perceptron for classification; and (4) feedback mechanism. The system has been preliminary validated by using data available from Japanese Female Facial Expression (JAFFE) database. Four affective states were identified and tested, which includes anger, sadness, and happiness. Subsequent tests have shown the successful detection rate of 91.3% with test images, and over 70% correct classification in images with Gaussian noises, respectively.