Optimal Identification and Estimation in Practical Noisy Environments

The practical environment settings are always dynamic and unpredictable. And the adaptive solution is deteriorated seriously unless we can accurately determine when the filter´s adaptation actually converges. In this paper, based on the principle of orthogonality, a scheme is proposed for sharply judging the iteration´s convergence to obtain the optimal estimation in practically unknown stationary or nonstationary circumstances. The discriminant obtained through the estimated mean-square values of the desired, output and error signals at each iteration cycle, can be updated according to the varying characteristics of the actual input signal. Cases of both computer simulations and real applications are studied to validate its effectiveness in stationary and nonstationary environments