Optimization Algorithm for Driver Monitoring System using Deep Learning Approach

The driver monitoring system (DMS), also known as driver attention monitor, plays an important role for vehicle safety systems. In DMS, the system detects the driver's activities such as the state of being a driver sleepy or failure to give sufficient attention to avoid accidents. To reduce the driver mistakes while driving, the system warns the driver with an alarm sound or vibrations. For the efficient implementation of DMS, the system should work in real time without any delay. However, the current DMS has many challenges implementation to the complexity of network implementation. To reduce the DMS complexity for real time implementation, we propose an optimization algorithm, which uses a camera images for monitoring the driver activities. From the input camera images, we extract the driver's state information from the region of interest (ROI). In addition, the proposed system also extracts the driver's head pose and gaze information and monitors the driver states during driving. The experiment results from the proposed methods show accurate driver's state information and warns the driver immediately when any mistake occur from driver side.