VoIP Voice and Fax Signal Processing

6.7 ADAPTATION FILTERING ALGORITHMS

The echo path consists of a linear and a nonlinear part. The linear part is modeled as a finite impulse response (FIR) or a moving average (MA) filter. G.168 [ITU-T-G.168 (2004)] has eight echo path models that work exactly as FIR filters. The nonlinear part requires higher order filter models. Volterra filters [(Mathews et al. (2000), Borys (2001)] model higher order nonlinear terms. In the literature, the applicability of Volterra filters were proved to be suitable for echo cancellation [Borys (2001)], but the utility is limited because of the higher memory and processing requirements of the Voltera filters. Nonlinear echo cancellation with higher order filters is beyond the scope of this book.

Linear part adaptation is estimated using popular least mean square (LMS) [Haykin and Widrow (2003), Haykin (1996)] and recursive least squares (RLS) based algorithms. LMS was widely adapted in the early designs of the adaptive filters. LMS algorithms are simple and stable as well as consume less memory and processing. These filters are updated in either the frequency domain or the time domain. If the filter span is longer, frequency-domain techniques are advantageous. The main disadvantage of LMS algorithms are that they adapt slowly. However, these algorithms were able to achieve the G.168 requirements. LMS algorithm adaptation further slows down with low signal levels. Normalized LMS (NLMS) normalizes signals that improve adaptation even for low-volume ...