ACOUSTIC ECHO CANCELLATION MENGGUNAKAN ALGORITMA NLMS

Abstract

Conversation that occurs in phone device usually almost unclear to be heard even though voice quality is the most important things in sending information process. One of problems that disturbing the conversation is something that we called echo. Echo is a phenomenon where a sound is reflected by a reflector object such as wall, floor, etc and it is also take some delay time. In hands-free communication system, at source side, the acoustic echo can be occurred because the voice from loudspeaker also get into microphone so that the information signal that has been sent will be mixed with this echo. Acoustic Echo Cancellation (AEC) is a method that will be implemented to clear this echo. AEC is built using adaptive filter with FIR filter structure. Echo is modeled as convolution result from input signal and room transfer function. Error signal that produced from this process will become weight component of NLMS adaptive algorithm to adapt the adaptive filter coefficient in order to minimize Mean Square Error (MSE). The result of AEC simulation test in stepsize (μ) = 0.7 and filter length (L) = 1024 is 1.07395 x 10-5 for the best MSE value and 30.6194 dB ERLE value as the optimum result when using Mahmud voice in 1b type room. The lowest value for Ucik voice in 3b type room, the result is 4.25501 x 10-5 MSE value and 16.6402 dB ERLE value. Keywords : echo, hands-free, adaptive filter, NLMS, MSE, ERLE