Open AccessApplication of blind source separation in sound source separation
Author(s) -
Jiacheng Xu
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1345/3/032006
Subject(s) - independent component analysis , computer science , source separation , blind signal separation , audio signal , speech recognition , sorting , signal (programming language) , sound (geography) , acoustics , artificial intelligence , algorithm , speech coding , telecommunications , channel (broadcasting) , physics , programming language
The classic method for solving the cocktail party problem is by utilizing Independent Component Analysis (ICA) method to separate different sounds. Since ICA method deals with each frequency point of the audio signal individually, there is a classic sorting problem. However, for Independent Vector Analysis (IVA), which makes use of the correlations between different frequency points of the same sound source, the audio of the same sound source can be isolated at one time without sorting problems. For music signal, due to the rhythm of music, there may be a strong correlation between different frequency points of the same sound source, so IVA method could be used for signal separation. This paper will discuss the application of IVA method in music sound separation and its application in music related speech recognition system from three aspects: basic principle, code implementation and performance analysis.