Low-Power Listen Based Driver Drowsiness Detection System Using Smartwatch

Abstract

Drowsy driving is a major cause of car accidents, because drivers are unable to swiftly perceive, process, and respond to the varying road conditions. Existing detecting solutions includes checking eye-blink, monitoring heartbeat with EEG or ECG device support, and analyzing the way the driver steers the steering wheel. Though effective, these solutions require extra hardware which causes distraction and inconvenience to the driver. We design and implement an unobtrusive and energy-efficient driver drowsiness detection system using only a commercial smartwatch through monitoring the steering behavior and heart rate of the driver. The system comprises two major modules, a hand state monitor and a drowsiness detector. The system is built by following insights. First, when the hand wearing smartwatch is off the steering wheel, no validate steering data will be captured. Thus it’s necessary to detect whether the hand is on the steering wheel to ensure the validity of the steering motion data. Second, heart rate features can reflect the alert level of the driver, and it can work no matter whether the hand is on the steering wheel. Consequently, we adopt the heart rate sensor of the smartwatch as a supplementary indicator of driver’s drowsiness level. Meanwhile, power consumption is considered given the limited smartwatch battery power. We evaluate our drowsiness detection system using a driving simulator, and it achieves an accuracy of 94.39%.