A Novel Variable Step Size LMS Algorithm Based On Modified Hyperbolic Tangent and Its Simulation

In the paper, a new variable step size LMS algorithm based on modified hyperbolic tangent is presented. In the algorithm, the step size is adjusted by the estimation of the autocorrelation between e(n) and e(n - 1). The algorithm introduces the compensation monomial to improve the convergence and the parameters to improve the shape and bottom characteristic of hyperbolic tangent. Therefore, the algorithm has faster convergence, better performance of noise suppression, lower steady state error and misadjustment. The theoretical analysis and simulation results all show that the overall performance of the new algorithm exceeds greatly some existent others under low SNR condition.