Drowsiness prognosis using chaos theory

Drowsiness is a state of impaired awareness or decreased consciousness related to a desire or inclination to sleep and difficulty in remaining alert [1]. It is considered one of the leading causes of truck accidents in the mining industry, causing irrecoverable economic, health, and life losses. An intelligent prognosis system can help the mining industry save the operator's life and expensive mining instruments. Despite the significant progress in drowsiness detection in recent years, the reliable early prognosis of drowsiness is still challenging. Electrocardiogram (EEG) base drowsiness detection method is a reliable approach that may be implemented by applying wearable devices [3]. This research aims to discover accurate fractal dimension and entropy algorithms that can be applied to EEG signals to compute reliable and effective indices for early drowsiness prognosis. Our approach takes advantage of chaotic quantifiers, including fractal dimension and entropy indices for feature extraction from EEG signal during the alert to a drowsy state transition. To accomplish this, a thorough analysis and evaluation were undertaken to examine the sensitivity and robustness of chaotic indicators, which included five fraction dimension algorithms and four main entropy approaches in terms of their capacity to forecast early drowsiness. According to the extracted feature evaluation, Higuchi and Katz's fraction dimension, Fuzzy and Permutation entropy indices perform better in discriminating alert and drowsy states. In this study, we utilized the fusion of different indicators for the proposed classifier. We trained and tested an SVM classifier that provided high performance by selecting a compact set of features that offer the greatest differentiability between the alert and drowsy states. Experiment results reveal that based on four fractal dimensions and entropy fusion, our strategy improves classification performance in distinguishing between the alert and drowsy states, with an accuracy of 96.30%.