A two step size NLMS adaptive filter algorithm

A new FIR adaptive filter algorithm is proposed. In the proposed algorithm the step size adjustment is controlled by using the square of the output error. A large output error will cause the step size α ₂ to provide faster tracking, while a small-output error will result in a step size α₁ yielding a smaller misadjustment. If when the mean square error is larger than some boundary, the, step size is set equal to α₂ and when the mean square error becomes smaller than the boundary the step size is set equal to α₁. There are two ways to select the umbral decision; by a fixed umbral and by a correlation criterion. Computer simulations show that the proposed algorithm achieves a better mean square error than a normalized LMS algorithm. the authors also compared the proposed algorithm with another variable step algorithm, achieving a better convergence time with the same mean square error