Bio-inspired wearable computing architecture and physiological signal processing for on-road stress monitoring

This work aims at developing a safety-critical wearable computer capable of monitoring the affective state and incremental changes in stress level of automotive drivers. This paper presents a perspective on the system design requirements and trade-offs, architectural aspects, affective state diagnosis and associated physiological signal processing for wearable driver assist systems. The proposed system architecture constitutes seven inter-connected functional units viz. Signal Acquisition and Conditioning Unit, Body Worn Unit, Signal Processing Unit, Human Intervention Unit, Local Computing and Actuating Unit, Back-End Computing Unit and Storage Unit. For classifier design, the data collected under five carefully designed scenarios is used to train the layer recurrent neural network for affective state recognition. We achieved an overall prediction rate of 89.18%, sensitivity of 89.63% and specificity of 94.17%. Finally, a semi-parametric Cox Proportional Hazard model was developed for estimating the incremental stress levels and detecting alarmable situations during on-road driving.