Real-Time Driver Depression Monitoring for Accident Prevention in Smart Vehicles

The alarming rise in road accidents underscores the urgent need for safer vehicles equipped with technologies to mitigate accidents and resulting injuries. Research in this area must consider driver fatigue and depression as they are among significant contributing factors to road accidents. This paper proposes a robust method for identifying depression in drivers using facial expression recognition technology. We leverage transfer learning, employing the VGG-16 model, for depression detection in drivers. The approach benefits from the pre-trained model’s knowledge, enabling it to adapt quickly to the task and reduces the data required for training. A dataset of images depicting normal and depressed drivers is utilized for training and testing the proposed model. The trained model analyzes a driver’s facial expressions to detect behavioral changes associated with depression. Upon detecting signs of depression, the system enables initiating a safe transfer of control from the driver to the vehicle’s automated systems. This automatic intervention aims to reduce the risk of accidents arising from impaired driver behavior. The proposed model is evaluated using various performance metrics achieving an accuracy of 96%. It also demonstrates excellent precision (98%), recall (97%), and F1-score (97%).