Multiple short-length adaptive filters for time-varying echo cancellations

An adaptive block selection algorithm for multiple echo cancellation is presented. The adaptive structure contains two echo path models. The primary model is an adaptive delay filter which is used to identify the most significant delays in the sparse system. The secondary model is a filter bank composed of a series of short length transversal filters which are centered around the active regions of the impulse response. An auxiliary component of the adaptive structure is a monitor/control algorithm which tracks time-varying echoes. This auxiliary algorithm positions the short length fibers so that they operate in the center of the active regions. Taking advantage of the sparse characteristics of the echo, the block selection algorithm reduces the computational effect of adjusting taps that would otherwise converge to zero, consequently increasing the convergence rate and decreasing the misadjustment for the overall system.<>