PEMISAHAN SINYAL AUDIO TERCAMPUR SECARA OFFLINE DENGAN ALGORITMA INDEPENDENT COMPONENT ANALYSIS

Abstract

Mixed audio signals can be found in a room where some people are speaking simultaneously, such as cocktail party or meeting. Someone’s ears will process the mixed signals automatically to get the meant audio signal, so the person can understand the message from the other. The mixed signals have random variables which aren’t determined before. The mixed signals separation method is called Blind Source Separation (BSS). Independent Component Analysis is a method (algorithm) for performing BSS. The principle of ICA is to estimate un-noised audio signals (closely related to the original signals) from the available mixed signals. A simulation program is needed to execute ICA in offline way. The program in this paper uses Matlab 7.1. It produces estimated audio signals and MSE (Mean Square Error) value between the estimated and original signals. The result of test indicates that FastICA algorithm is the eficient method to compute parameters quickly to get the right estimated signals. The most efective model of this algorithm is symmetric approach. Kata kunci: audio, offline, FastICA, MSE