Enhanced frame rate for real-time eye tracking using circular hough transform

Eye-gaze detection and tracking has been widely investigated and presented as a way of unconventional human computer interaction. This area has provided convenience for many fields of practical applications, such as assistive systems and technology for people with severe disabilities, virtual reality, driver assistance and monitoring systems. Many methods for eye tracking have been introduced in literature. In this paper, a real-time eye tracking system is presented. To locate the iris of the eye in the captured video frames, the system uses the Circular Hough Transform which aims to recognize circular patterns in an image. Generally, the speed of eye motion is not as high as the used video frame rate of 30 Frames per Second (FPS) which is the frame rate used in general live video. In other words, the eye cannot move as fast as 30 motions per second. This led to proposing an enhancement to the eye tracking system being presented. This enhancement improved the CPU processing time requirements. The enhancement presented in this paper suggests that not all captured live video frames need to be processed for eye detection because the same eye movement will be captured on multiple subsequent video frames. Processing only a subset of frames will be enough to detect all eye movements in the video. The required CPU processing time is improved by selecting the minimum accepted video frame rate sufficient for accurately detecting all eye motions in a video. This was investigated for both the low and high speed eye movements. For low speed eye movements, the improvement in required CPU time was 1500%. For high speed eye movements, it was 750%. The improvement in CPU time is general and applies to different eye tracking algorithms when using the proposed enhancement. These improvements are a result of the elimination of the redundant video frames which are no longer processed in the procedure of eye detection.