BROWSE
Detail
Conference Paper
Convolutive Blind Separation based on Warped Linear Prediction Analysis
Cited - time in 
Share 
Share
- Issue Date
- 2006-12
- Citation
- International Congress and Exposition on Noise Control Engineering (INTER-NOISE) 2006, pp.5693-5699
- Language
- English
- Type
- Conference Paper
- Abstract
- Convolutive source separation (CBS) is a technique to restore original unknown sources from multiple microphone signals that are modeled as convolutive mixtures of mutually independent sources. In this paper, we propose a CBS algorithm, which is based on multichannel blind deconvolution (MBD) method with warped linear prediction (WLP) analysis to overcome the side effect, undesired temporal whiteness of separated signals of MBD. Through the experiments of real recording data, we showed the high performance of our proposed algorithm.
- KSP Keywords
- Blind Deconvolution, Blind separation, CBS algorithm, Convolutive Source Separation, Convolutive mixtures, High performance, Linear prediction analysis, Side effects, Warped linear prediction