Aplikasi Metode Independent Component Analysis untuk Pemisahan Sinyal Wicara dengan Backsound pada Audio Movie
DOI:
https://doi.org/10.36277/jteuniba.v7i2.209Abstract
Abstract: Voice analysis in establishing speech and non-speech is carried out by audio signal processing from the sound source. This research focuses on mixed voice data in the form of actor voices (speech signal) and background voices (backsound) from a movie. In this study, sound signal separation has been carried out using the Blind Source Separation (BSS) method to separate mixed signals into a number of forming signals without information about the number of signal sources or the process of mixing these signals. The algorithm used for BSS in this study is the Independent Component Analysis (ICA) algorithm. The actor's voice signal that has been separated from the background sound in a film was tested using MSE and SIR analysis to determine the quality of the separation signal. From the Mean Square Error (MSE) test of the speech signal from the separation results, a test value of 5 seconds was obtained of 0.036, for a duration of 10 seconds the MSE value was 0.0432, and in the data of 15 seconds it produced an MSE value of 0.0558. Signal to Interference Ratio (SIR) analysis of speech signals resulting from separation, for data with a duration of 5 seconds of 14,485, for data measuring 10 seconds obtained SIR of 13,645, and for data with a duration of 15 seconds obtained SIR of 12,533. The output of this sorting process is then used as an input for other necessary signal processing processes.
Intisari— Analisis suara dalam menetapkan speech dan non-speech dilakukan dengan audio signal processing dari sumber suara. Penelitian ini berfokus pada data suara campuran berupa suara aktor (speech signal) dan suara latar (backsound) dari sebuah film. Pada penelitian ini telah dilakukan pemisahan sinyal suara dengan menggunakan metode Blind Source Separation (BSS) untuk memisahkan sinyal tercampur menjadi sejumlah sinyal pembentuknya tanpa informasi mengenai jumlah sumber sinyal atau proses tercampurnya sinyal-sinyal tersebut. Algoritma yang digunakan untuk BSS dalam penelitian ini adalah algoritma Independent Component Analysis (ICA). Sinyal suara aktor yang telah terpisah dari suara latar pada sebuah film diuji menggunakan analisa MSE dan SIR untuk mengetahui kualitas sinyal hasil pemisahan. Dari pengujian Mean Square Error (MSE) terhadap sinyal speech hasil pemisahan, diperoleh nilai pengujian pada durasi 5 detik sebesar 0.036, untuk durasi 10 detik nilai MSE yakni 0.0432, dan pada data sebesar 15 detik menghasilkan nilai MSE sebesar 0.0558. Analisa Signal to Interference Ratio (SIR) terhadap sinyal wicara hasil pemisahan, untuk data berdurasi 5 detik sebesar 14.485, untuk data berukuran 10 detik diperoleh SIR sebesar 13.645, dan untuk data dengan berdurasi 15 detik diperoleh SIR sebesar 12.533. Luaran dari proses pemilahan ini selanjutnya digunakan sebagai input bagi proses pengolahan sinyal lainnya yang diperlukan.
Downloads
References
C.Y. Fook, M. Hariharan, S. Yaacob, and A. Ah, “A Review : Malay Speech Recognition and Audio Visual Speech Recognition,” no. February,pp. 27-28, Universiti Malaysia Perlis (UniMAP), 2012.
M.V. Gubin,” Using Convolutional Neural Networks to Classify Audio Signal in Noisy Sound Scenes”, 2018 Global Smart Industry Conference (GloSIC).
Snehal S. Gaikwad, Pallavi P. Ingale, Dr. S. L. Nalbalwar, “Separation of singing voice from background musical noise using modified NMF and Filtering”, International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT) - 2016 .
Pritish Chandna, Merlijn Blaauw, Jordi Bonada, Emilia Gomez, , “A Vocoder based Method for Singing Voice Extraction”, ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).
Kenneth John Faller, Jason Riddley, and Elijah Grubbs, “Automatic Blind Source Separation of Speech Source in an Auditory Scene ”, 2017 51-st Asilomar Conference on Signals, Systems, and Computers.
Yang Yang, Zuoli Li, Xiuqin Wang and Di Zhang, “Noise Source Separation based on the Blind Source Separation”, 2011 Chinese Control and Decision Conference (CCDC), Mianyang, China ,23-25 May 2011.
R.Farkhan, “Application of blind source separation technique for separating noise from non gaussian acoustic signal”, Institut Teknologi Sepuluh Nopember Surabaya, 2013.
P. Angga Pramana, W. Ni Wayan, M. Ni Putu, B. I Dewa Made, “Independent Component Analysis (ICA) dan Sparse Component Analysis (SCA) dalam Pemisahan Vokal dan Instrumen pada Seni Geguntangan”, Jurnal Elektronik Ilmu Komputer Udayana p-ISSN: 2301-5373 Volume 8, No 1, Universitas Udayana, 2019.
Houssem Maghrebi and Emmanuel Prouff, “On the Use of Independent Component Analysis to Denoise Side-Channel Measurements”, International Workshop on Constructive Side-Channel Analysis and Secure Design, COSADE 2018. In the book of Constructive Side-Channel Analysis and Secure Design pp 61–81.
M. Shoji, S. Hiroshi, M. Ryo, A. Shoko, “Blind Source Separation of Convolutive Mixtures of Speech in Frequency Domain”, IEICE Trans. Fundamental, VOL.E88-A, NO.7 July 2005.
T. W. Lee. “Independent Component Analysis”, Kluwer Academic Publishers, Boston, 2000.
T.Sihotang, A.Asni, dan Anwar Fattah, ” Ektraksi Ciri Menggunakan Algorithma Discrete Wavelet Transform (DWT) dan Principal Component Analysis Pada Warna Kulit wajah”, JTE UNIBA, Vol. 3, No. 2, April 2019.
Saftiadi, A. Asni B, Aswadul Fitri Saiful Rahmant, “ Perancangan Sistem Kontroler Alat Elektronik Rumah Tangga berbasis Miklrokontroler Arduino dengan Perintah Suara”, JTE UNIBA. Vol. 3. No. 2, April 2019.
A.Asni B, Diah Patriana Setianingsih, “ Pengenalan IndentitasPenutur MenggunakanAlgoritmaDiscreteWavelet Transform (DWT)dan Hidden Markov Modesls (HMM)”, JTE UNIBA, Vol. 05, No 1. September 2018.
Downloads
Published
Issue
Section
License
Copyright (c) 2023 Titon Dutono, Dinda Ayu Oktaviasari, Tri Budi Santoso
This work is licensed under a Creative Commons Attribution 4.0 International License.
