Drowsiness Detection and Classification System Design: Towards Better Road Safety

Abstract

The figure of road accidents is rapidly growing across the world. Among many other factors, driver's drowsiness is one of the prime sources of such crashes. In drowsy state, the decision taking capability and alertness level of the human brain decreases which results in the accident. In order to control such accidents, the drowsy condition of the driver has to be identified in advance. An Electroencephalogram (EEG) based drowsiness monitoring and alarming system by extracting DWT (Discrete Wavelet Transform) features for higher accuracy is presented here. The scope of the paper is to extend the concept into a smart IoT device with ability to generate alert and classify drowsy state of driver using wavelet transform. The system is tested on 60 subjects in both awake and drowsy states by applying ANN (Artificial Neural Network) and KNN (K Nearest Neighbour) algorithms. Using six training functions on ANN algorithms, accuracy of 90%, sensitivity of 95% and specificity of 90% is achieved. Similarly, the best results from KNN algorithm are the accuracy of 94.92%, sensitivity of 93.10% and specificity of 94.92%.