Wearable Driver Drowsiness Detection System in Smartwatch

Abstract

Driver drowsiness detection system had been developed as comprehensive application but this has the potential risk of distracting the driver’s attention, causing accidents. Thus, a wearable-type drowsiness detection system is proposed to overcome such issue. The proposed system used self-designed wristband consisted of photoplethysmogram sensor and galvanic skin response sensor. The sensors data are sent to the smartwatch which served as a main analyzing processing unit. Those data are analyzed along with the motion data collected from built-in accelerometer and gyroscope sensors. Features extracted from the data based on data time-domain, frequency domain, and phase domain. The correlation of each feature was calculated using PEARSON correlation method. Those top features are further served as computation parameters to a support vector machine (SVM) to derive the driver drowsiness state. The accuracy of system had calculated based on number of input features into SVM. The testing results indicated that the accuracy of the system with SVM model reached up to 98.15% with 8 features. If drowsy detected, driver will be alerted using graphical and vibration alarm generated by the smartwatch. In fact, the integration of driver physical behavior and physiological signals is proven to be an outstanding solution to detect driver drowsiness in a safer, more flexible and portable used.