Echo cancellation in Voice over IP

Abstract

The emerging technology called Voice over IP has been widely recognized as the candidate system against the PSTN in voice communication. Nonetheless, the quality of signal transmission over IP is still a major concern because of the echo problem. Echo is one of the serious types of noise commonly found in VoIP communication. In general, there are two types of echo problems in VoIP, hybrid/network echo and acoustic echo. In this paper, an approach to echo cancellation in VoIP network based on the packet-based echo cancellation model is employed. In addition, we also analyze and perform comparative validation specifically in the state of echo canceller with adaptive filtering as done by using two class measurements: mean square error (MSE), and least square (LS) methods. For experimentation, algorithms of normalized least mean square (NLMS), recursive least square (RLS), and the proposed least square lattice (LSL) are applied in the echo canceller for comparison. Communication environments and three different echo signal strength levels: 20%, 50%, and 70% of far-end speech are simulated. The result shows that in terms of computational time, the NLMS is the best among all, followed by the LSL as the second. However, the NLMS suffers from the measure of ERLE, as it falls to the third place, followed by the LSL as the second, and the RLS as the first. Then, taking them all into consideration, we could conclude that the proposed LSL is the best method among all for echo cancellation in VoIP communication.