Enhancing echo cancellation via estimation of delay

Abstract

The advent of packetized audio transmission, such as voice over IP (VoIP), has resulted in challenging requirements for echo cancellation technology. One key aspect of this technology is the need to characterize, quickly and accurately, the echo paths in the transmission media. Echo paths consist of a constant time delay with no echo signal and active regions in which the echo signal is present. When an adaptive filter echo cancellation algorithm is used, its performance can be greatly increased, and its complexity can be reduced if it is only applied to the active regions. This requires an algorithm to estimate the constant delay and locate the active regions. Traditionally, delay estimation has been based on direct application of cross-correlation. This method has poor performance because the input signals are highly correlated and has a high implementation cost because many cross-correlation lags have to be computed for longer time delays. The delay estimation addressed in this paper has two major advantages over the traditional methods. The first is that it has improved performance because the input signals are processed to have less correlation. The second is that the implementation cost is significantly reduced because fewer cross-correlation lags are computed, and an efficient method to estimate lags is created.