<p>Drowsy driving is a significant risk worldwide, leading to accidents and loss of life. The present study explores histogram of oriented gradients (HOG) and support vector machines (SVM) as a prominent approach for identifying the driver drowsy state via facial recognition technique in real time. The facial recognition system was built on HOG and SVM, along with its components, functionality, and effectiveness. The Python program was used as the backend language using Dlib and OpenCV. To enhance real-time accuracy, the system continuously monitors subtle changes in facial expressions, such as slow eye blinks and changes in head position. These micro-expressions, often overlooked in conventional systems, act as early indicators of fatigue. The proposed system is designed to work even in scenarios where the driver wears glasses or a mask, making it adaptable to practical driving conditions. The driver's condition was categorized into three subcategories i.e., tired, dynamic, and resting. The real-time input information of the driver's state was recognized using the 68 facial landmark detectors along with the position of the face and advanced focus on other facial features. The system dynamically adjusts its detection thresholds based on lighting conditions, ensuring consistent performance during day and night driving. The different ratios were used to predict whether the driver is in an active, lethargic, or languid state. The findings of the study show that HOG and SVM have very high accuracy rates (96.8%) in ideal conditions. Moreover, the proposed approach demonstrated resilience to minor facial occlusions, which is a common challenge in real-world deployments. The study can be helpful in designing and developing early warning systems to detect signs of drowsiness and subsequent efforts to avoid accidents and loss of life. The proposed work will be a milestone in improving safety in case of drowsiness, facilitating reducing the number of accidents caused by drowsy driving.</p>