Application of a new OLS learning algorithm for video ghost cancellation

Very successful multiple video ghost cancellation simulations have been obtained with the application of a designed learning algorithm, based on the orthogonal least square method, to a channel identification process based on a FIR system model. The algorithm can be visualized by assuming the existence of a data matrix, at the input of the equalizer, which is created with a time shifting process, in order to generate “M” column vectors. The designed algorithm processes these vectors and, with the aid of an orthogonalization method, calculates a set of the most representative one, with respect to a desired output signal. The ghost parameters were obtained with the aid of a forward regressor method. The process has shown to be very effective for the accurate calculation of the ghost parameters, even in the presence of considerable noise levels, and is also used to train RBF approximation networks for the systematic selection of its centroids. In all the tests performed in this work, the proposed technique has given much better results than using conventional algorithms