A Study on Source Separation Using Orthogonality between Independent Speeches in Spectrogram

Author

Jang, Hyukjoon ; Jeong, Hong

Author_Institution

Dept. of Info. Technol., POSTECH, Pohang, South Korea

fYear

2010

fDate

9-10 Jan. 2010

Firstpage

338

Lastpage

341

Abstract

We propose a kind of source separating method in audio signal. It uses an orthogonal property in spectral domain, which is an extended concept of temporal orthogonality used in Degenerate Unmixing Estimation Technique (DUET) to individual frequency bands. In addition, we also adopt a gradient-based learning rule using negentropy for perfect separation. In this paper, we will describe our proposed algorithm and show experimental results.

Keywords

audio signal processing; gradient methods; independent component analysis; learning (artificial intelligence); source separation; speech processing; audio signal; degenerate unmixing estimation technique; gradient based learning rule; independent speeches; negentropy; orthogonal property; source separation method; spectrogram; temporal orthogonality; Data mining; Frequency estimation; Histograms; Image converters; Independent component analysis; Source separation; Spectrogram; Speech; Time domain analysis; Time frequency analysis; DUET; FD-ICA; NBF; Orthogonality;

fLanguage

English

Publisher

ieee

Conference_Titel

Knowledge Discovery and Data Mining, 2010. WKDD '10. Third International Conference on

Conference_Location

Phuket

Print_ISBN

978-1-4244-5397-9

Electronic_ISBN

978-1-4244-5398-6

Type

conf

DOI

10.1109/WKDD.2010.22

Filename

5432601