Abstract:
The work presented here is part of a longterm research project that aims at creation of a wearable driver assistance system WDAS that could be used to prevent loss of lives and fatal injuries which may be caused due to road accidents In particular this work focuses on the realtime acquisition and analysis of physiological signals noninvasively sensed from automotive drivers by making use of bodymounted sensors Realtime data acquired in this way could be used for timely detection of physiological state of the driver that may otherwise lead to unsafe driving newlineMethodology used included identification of an exhaustive set of features or attributes which as per literature and collectable primary data could lead to determination of the most significant parameters which meaningfully and credibly indicate affective state of a driver Using a set of shortlisted parameters like Heart Rate HR Heart Rate Variability HR Skin Conductance SC level blood oxygen saturation also known as Saturation of peripheral Oxygen SpO2 and respiration rate etc a set of realtime data collection experiments were designed to provide the primary data for the purpose of this research As a consequence for over a year several experiments were conducted on different drivers in predriving indriving and postdriving states with appropriate sensors mounted on their body with their consent In the next stage the data collected in such a manner was cleaned duly formatted and thereafter subjected to appropriate methods of analysis The entire process resulted in not only extraction of appropriate feature sets but also identification of a very small subset of parameters realtime sensing of which would allow creation of resultant architectural framework that would pave the way for actually building a costeffective and robust wearable computing system for the vehicular drivers newlineIn this context a driverprofile analysis based on the Cox Proportional Hazard model firmly established that the Current Physiological State CPS was the most important predictor with highes
