Abstract
Drowsy driving is one of the major causes of traffic accidents, especially for truck drivers. In this article, we propose a novel algorithm for drowsiness detection based on driving signals. The steering velocity is filtered by a weighted discriminative frequency response that has a magnitude response for each frequency bin proportional to the Kullback-Leibler divergence between two drowsiness states (“awake” and “drowsy”). Such frequency responses can be obtained off-line from the training data. A comparison of the cumulative sum of the filtered steering velocity and an adaptive threshold based on a lateral preview driver control model detects the driver’s drowsiness state. The proposed framework was compared to a conventional algorithm using corpuses on a Hino Motors test course and on expressways. Results showed that the performance of the proposed framework was 7% better than the conventional one in terms of equal error rate (EER).
Keywords: Driver State Detection, Driver Assistance System