Enhancement of an Echo Cancellation Method

Abstract

In modern hands-free telecommunication and teleconferencing systems, during newlineconversation the main problem is the occurrence of an acoustic echo. An acoustic echo causes newlinedistraction which affects the quality of speech in communication. An electrical signal or original newlinesound when delayed and distorted is reflected back to the source then that sound or signal is newlinecalled as an echo. In hands free communication instead of a telephone receiver, a high-gain newlinemicrophone and a regular loudspeaker is used which is more suitable for communication. As in newlineteleconference environment this allows more than one person to communicate with each other newlinesimultaneously. Most important advantage here is it allows the person to move freely in the room newlinehaving his both hands free. However, a loud echo is produced due to a large acoustic coupling newlinebetween the microphone and loudspeaker which makes the conversation of two persons difficult. newlineFurthermore, because of all these factors the instability of the acoustic system increases, which newlinecause lot of interference and noise to occur. The removal of echo using an efficient echo newlinecancellation algorithm is a perfect solution to these problems. In full duplex communication the newlinespeakers at both ends can talk simultaneously due to which echo cancellers can be easily newlineimplemented and maintained. The main aim of this proposed work is to enhance an algorithm newlineused for acoustic echo cancellation, which gives satisfactory results. The adaptive algorithms are newlinemainly used to achieve higher ERLE, fast convergence rate with low complexity of newlinecomputations. An adaptive filter and a doubletalk detector are the two basic components of an newlineecho canceller. The output of adaptive filter is subtracted from the microphone signal. The newlinesimultaneous talk from both ends is sensed by the double talk detector. When speaker at both newlineends are talking simultaneously, the double talk occurs due to which there are chances of newlineadaptive filter to diverge. To avoid this adaptive filter stops adapting its coefficients a