Facial tracking using artificial neural networks

A region-based method for facial tracking is proposed in this paper. In this method, the facial information of temporal motion and spatial luminance are fully utilized. The dominant motion of the tracked facial object is computed. Using this result, the object template is warped to generate a prediction template. A method is proposed, which incorporates an artificial neural network (ANN), with back-propagation algorithm (BPA), to modify the prediction. A decision approach, with a threshold, is used to detect, if there is any change in the object of the successive frames. The accuracy of the result depends upon the number of nodes in the hidden layer and learning factor. The number of nodes in the hidden layer is 10, and the learning factor is 1. The performance of the algorithm, in reconstructing the tracked object, is about 96.5%, in terms of reduced time and quality of reconstruction.