ECG based driver drowsiness detection using scalograms and convolutional neural networks

Driving while drowsy is one of the significant causes road accidents, and detection drowsiness drivers necessary to minimize such accidents. There are many methods proposed for detection. In this paper, we present details a system using deep learning techniques with ECG as its input. Continuous wavelet transforms (CWT) applied on input signal, segmented 1-minute duration, then it converted into scalogram images be fed neural networks (DNN). We use pre-trained transfer models, AlexNet, ResNet-50, VGG-19 work. The best results were obtained Reset-50 an accuracy 88.31% without dropping final layers, which improvement 4.5% over spectrogram-based implementation. This paper evaluates effectiveness wavelet-based features, note that time-frequency domain features outperform. To check robustness net-work, second model implemented by validation AlexNet 51.5%, ResNet-50 67.2%, 81.8%.