Enhanced Real-Time Driver Fatigue Detection Using Hybrid SVM and Multi-Feature Analysis for Improved Accuracy and Road Safety

Fatigue and distraction are two of the main causes of traffic accidents, so it's critical to maintain focused attention when driving. A ‎reliable real-time fatigue detection system based on driver visual ‎analysis is proposed in the present research. Yet, features like interference from eyewear and low illumination can affect the precision ‎of detection. A machine learning-based technique for detecting and ‎classifying driver fatigue in real-time is presented in this research. ‎Along with blinking, additional indicators of fatigue include yawning, nodding, and variances in pupil size. To extract multiple visual ‎cues, the Image Projection Function (IPF) is used to quantify pupil ‎size, optical flow is employed for assessing head position, and eye ‎and mouth dimension ratios are measured. Normalization is per-‎formed to ensure data consistency. Multiple facial fatigue ‎markers are incorporated into the system to improve classification ‎accuracy and robustness. Principal Component Analysis (PCA) is ‎applied for dimensionality reduction to optimize classification performance. The Euclidean distance gauge is used to evaluate three ‎machine learning models: Support Vector Machine (SVM), Multi-layer Perceptron (MLP), and K-Nearest Neighbors (KNN). Outperforming KNN and MLP, SVM with the Radial Basis Function ‎‎(RBF) kernel obtains the best accuracy of 91.6% among them. The ‎classification performance is enhanced by the RBF kernel's ability ‎to capture non-linear fatigue-related patterns. The efficiency of the ‎proposed method for detecting fatigue in the real world has been ‎verified by experimental results, which indicate the attributes obtained ‎can reliably differentiate between fatigued and non-fatigued states‎.