A method for object tracking using user intention and dynamics of object area

摘要

In the source separation problem, knowledge of source locations improve the performance of separation. This is a motivation to develop our new hearing system which uses not only acoustical signals but also image sequences for source localization. The functions of the image processing sub-system of the hearing system are decision of the object region according to the intention of user and the real-time object tracking. In this paper, we propose a new method for object tracking which has the ability for these functions. In this method, at first, a location of a sound source and a shape of the template for image matching are estimated according to a pointing light located by user. Then, using image sequence, the location and the template are updated based on a statistical model; the updating induces the object tracking. Numerical simulations using synthesized image sequences including a object, changing its shape and location randomly, demonstrate robustness and efficiency of the proposed method by comparison with conventional template-matching method.