Humans can identify the relative position of a person and person’s movement. In computer vision, this is a complex problem. Face detection and recognition is at the forefront of face related vision research. In this paper, we perform face detection and track the face to move devices like camera or microphone. Our focus is to slide or rotate the devices such that they are constantly tracking the face. The videos recorded by a camera is streamed to the acquisition system. The first frame is extracted and used in detecting the face and ROI around the face is used as reference for tracking. Histogram for this ROI is computed and set as the reference histogram which is used in consecutive frames. The method employed for tracking is mean shift tracking with exponential kernel. The kernel with the initial histogram is used to perform back projection on the frame. For the consecutive frames, the peak in the kernel is located and is used for drawing the bounding box thereby tracking is performed and each frame is aligned with the face being the center of the frame, which is an indication for the devices to be moved according to the movements in the person.
The videos on the left of the page, we have the original recorded videos and we can observe that the person is moving back and forth. The videos on the right side of the page are generated after passing through our proposed method. We can observe that the camera is moving along with the person. This method finds application is several places like surveillance cameras and sports cameras.