Variable LMS Algorithms Using the Time Constant Concept

In this paper, two variable least mean square (LMS) algorithms using the simple time constant concept, variable step size LMS (VS-LMS) and variable length LMS (VL-LMS) algorithms, are proposed. In proposed variable LMS algorithms, the time constant by which the step size is changed in the VS-LMS algorithm and the filter length is varied in the VL-LMS algorithm, is used as a measure of convergence of the error signal. Computer simulation results show that proposed algorithms are simpler but show better performance than conventional variable LMS algorithms.