A new adaptive algorithm based on conditioned normalized LMS method

A new algorithm for updating the coefficients of an adaptive FIR digital filter is modified by using the preset value p (constant). This value is determined beforehand from knowledge of the probability density function and variance of the input signal. In the new conditioned NLMS algorithm, the measured input signal´s square norm is compared with the preset value p. If the square norm is larger than p the coefficients of the filter are renewed. No renewal of the ADF´s coefficients is done when the square norm is smaller than p. The simulation results and theoretical analyses show a good agreement. The saturated residual error in noisy circumstances is smaller than that of the unconditioned ordinary filter