LMS iterative algorithms applied to the computation of TV ghost canceller parameters

Three iterative forms of the LMS learning algorithm were tested for the calculation of the coefficients of FIR filters used as TV ghost cancellers. These computational forms are: the stochastic gradient fixed-step (SGLMS) and a variable step (SLS-CD), algorithms, as well as the recursive modified Gram-Schmidt RMGS algorithm. Because of the iterative nature of the selected algorithms, they are very convenient to be used in on-line LTF filter coefficient adaptations. This makes it possible to compute the coefficient values of the ghost canceller, when the sampling of the signal generates a huge amount of data, which is very hard to be handled with a PC. The aforementioned algorithms are written in a very powerful and flexible matrix oriented software, and all the tests were performed using a very flexible TV system simulator. During the tests, fast convergence of the ghost canceller coefficients to the theoretical values have been observed