PAFWF-EEGC Net: parallel adaptive feature weight fusion based on EEG-dynamic characteristics using channels neural network for driver drowsiness detection

Drowsy driving is considered one of the most dangerous causes of road accidents and deaths worldwide. Drivers’ concentration is directly affected by fatigue, which affects their reaction time, reducing their attention and decision-making ability on the road. This can often lead to dangerous situations. With the development of Human Computer Interface systems and the rise of intelligent transportation systems, examining the effects of driver fatigue has become more critical, and research aimed at reducing the risk of fatigue-related accidents has gained importance. For this purpose, this study proposes a Parallel Adaptive Feature Weight Fusion based on EEG-Dynamic Characteristics using Channels Neural Network (PAFWF-EEGC Net) to detect the driver drowsiness condition. Two signal processing techniques are used to extract EEG dynamic features: first, Continuous Wavelet Transform (CWT) to capture the spectral-temporal features by accurately estimating both time and frequency localizations, and second, Fast Fourier Transform (FFT)—Power Spectrum Density (PSD) to convert the signals from the time domain to the frequency domain and show the distribution of signal power over frequency. These extracted dynamic features are passed to Attention channels and Parallel Adaptive Feature Fusion to integrate the most relevant feature channels to detect mental state. Furthermore, three processing dataset scenarios and cross-validation techniques are used to validate the Net. The Net showed excellent performance through ninefold/3rd scenario by achieving 98% detection accuracy, and 84%, 88.75%, 93.8% average detection accuracy through 1st, 2nd, 3rd scenarios respectively.