IMPLEMENTASI DAN ANALISIS SISTEM PEMISAHAN SINYAL SUARA TERCAMPUR MENGGUNAKAN METODE BLIND SOURCE SEPARATION (BSS)

DWI, GIOVANNI (2019) IMPLEMENTASI DAN ANALISIS SISTEM PEMISAHAN SINYAL SUARA TERCAMPUR MENGGUNAKAN METODE BLIND SOURCE SEPARATION (BSS). Undergraduate Thesis thesis, Institut Telkom Purwokerto.

	Text A. Cover.pdf - Accepted Version Download (1MB)
Preview	Text B. Abstrak.pdf - Accepted Version Download (87kB) | Preview
Preview	Text C. Abstract.pdf - Accepted Version Download (33kB) | Preview
	Text D. BAB I.pdf - Accepted Version Download (112kB)
	Text E. BAB II.pdf - Accepted Version Restricted to Registered users only Download (283kB)
	Text F. BAB III.pdf - Accepted Version Restricted to Registered users only Download (305kB)
	Text G. BAB IV.pdf - Accepted Version Restricted to Registered users only Download (1MB)
	Text H. BAB V.pdf - Accepted Version Download (84kB)
Preview	Text I. Daftar Pustaka.pdf - Accepted Version Download (151kB) | Preview

Abstract

ABSTRACT The problem of speech signal mixture in communication system happens everyday. This problem is summarized as cocktail party, where more than one speaker speaks at the same time. This study aimed to implement speech signal mixture extraction from 2 independent sources using Independent Component Analysis (ICA). The system consists of sort of steps: 1) preprocessing which are centering and normalization; 2) Principal Component Analysis and Whitening; 3) FastICA with fixed point; and 4) performance evaluation using Signal to Noise Ratio (SNR) and Coefficient Correlation. There are 12 independent speeches that mixed up to get 48 new signals where each mixture comprised of 2 independent speeches. The result shows the lowest and highest SNR were 14,2156 dB and 85,1658 dB for first speeches signal group and 11,3501 dB and 75,4498 dB for second speeches signal group. The lowest SNR value at the first and second speeches signal group were a mixture of KASIHAN and PEPAYA. The highest SNR value at the first speeches signal group were a mixture of SURABAYA and MERASA where the highest SNR value lies in the word SURABAYA whereas at the second speeches signal group were a mixture of PURWOKERTO and KASIHAN MERASA where the highest SNR value lies in the word KASIHAN. Number of SNR values above 20 dB at the first and second speeches signal group were 44 and 46 or around 91,67 % and 95,83 %. While coefficient of correlation between corresponding extracted and independent speech signals are highly significant which were 0,9999 and 1,0000 for first and second signal consecutively. It implies that implemented system was able to separate mixtures of independent speeches signal. Keywords: Blind Source Separation, Independent Component Analysis, Sound Signal, Fast ICA, Principal Component Analysis

Actions (login required)

View Item