Several modified LMS algorithms are studied in order to improve the rate of convergence, increase the tracking performance and reduce the computational cost of the regular LMS algorithm. These methods can be divided in two categories: Clipped data algorithms and variable step size algorithms. In this paper a new quantized input variable step size LMS algorithm is introduced. The proposed algorithm is a modification of an existing method, namely, VSS LMS, and uses a new quantization function for clipping the input signal. We showed mathematically the convergence of the QX-VSS LMS filter weights to the optimum Wiener filter weights. Also, we proved that the proposed algorithm has better tracking than the conventional LMS algorithm. We discuss the conditions which one have to consider so that he can get better performance of QX-VSS LMS algorithm. The results of simulations confirm the presented mathematical analysis

Keywords