Performance of an adaptive echo canceller operating in a noisy, linear, time-invariant environment

In this paper we derive equations describing the performance of various adaptive echo canceller configurations operating in a linear, time-invariant environment. We relate the parameters in these equations to measurable environmental factors, discuss their effect on performance, and verify the results empirically. In general, the performance of an echo canceller cannot be exactly predicted for speech inputs. Therefore, the derived equations assume a stationary constant power random input. However, it is shown that the results obtained in this manner give useful estimates of the performance to be expected with speech inputs. The similarities and differences of the results for a constant power random input and speech input are discussed in detail.