Drivers’ Affect Recognition System Utilizing Multimodal Biosignals

Abstract

Estimating the affect of motorcycle drivers can improve safety measures during the drive. Previous researches have been attempting to predict people’s affect by employing the machine learning model and biosignals such as ECG and EEG. This paper proposes a novel real-time affect recognition method using multimodal biosignals employing knowledge distillation (KD) that enables the proposed model to estimate the driver’s affect using only ECG. An experiment involving 28 subjects was conducted to measure their biosignals when watching 360 videos with VR. Though the proposed model only utilized ECG during the test, experimental results demonstrated that it could achieve a satisfying performance.