Adaptive Filtering Algorithms With Automatic Gain Control

In this paper we introduce an automatic gain control (AGC) scheme for adaptive algorithms that are used extensively in many applications. The proposed AGC scheme is realized by using an estimate of the cross correlation between adaptive error and the input signal to control the gain of the adaptive algorithm. When the cross correlation is high, the gain is also high, and the adaptive algorithm is in an "active" state. When the error and the input signals are uncorrelated, the gain is near to zero, and the adaptive algorithm is put in an "asleep" state. Thus, adaptive algorithms with such AGC are insensitive to disturbances appearing on the system output measurement. Such disturbances can drive conventional adaptive algorithms away from the achieved adaptation. A fast, efficient algorithm for estimation of the cross correlation coefficient of adaptive error and input is also proposed.