CNN Based Subject-Independent Driver Emotion Recognition System Involving Physiological Signals for ADAS

Abstract

Supporting drivers by Advanced Driver Assistance Systems (ADAS) significantly increases road safety. Driver’s emotions recognition is a building block of advanced systems for monitoring the driver’s comfort and driving ergonomics additionally to driver’s fatigue and drowsiness forecasting. This paper presents an approach for driver emotions recognition involving a set of three physiological signals (Electrodermal Activity, Skin Temperature and the Electrocardiogram). Additionally, we propose a CNN (cellular neural network) based classifier to classify each signal into four emotional states. Moreover, the subject-independent classification results of all signals are fused using Dempster-Shafer evidence theory in order to obtain a more robust detection of the true emotional state. The new system is tested using the benchmarked MAHNOB HCI dataset and the results show a relatively high performance compared to existing competing algorithms from the recent relevant literature.