Loading...
Please wait, while we are loading the content...
Similar Documents
Méthode temps-fréquence de séparation aveugle de sources basée sur la fonction de cohérence segmentée
| Content Provider | Semantic Scholar |
|---|---|
| Author | Albouy, Benoit Deville, Yannick |
| Copyright Year | 2003 |
| Abstract | In this paper, we introduce a new blind source separation (BSS) method for linear instantaneous mixtures. This approach is based on the time-segmented frequency-dependent real coherence function of the observed signals, which makes it possible to detect time-frequency (TF) zones where only one source is active. In addition, this method only assumes the sources to be uncorrelated. The separating coefficient identification is based on ratios of power spectral densities of the mixed signals, computed in single-sources TF zones. This BSS method yields very high performance for mixtures of speech and/or noise signals, with signal/noise ratios improvements ranging from 40 dB to more than 90 dB and 100 % automatic speech recognition rates. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://documents.irevues.inist.fr/bitstream/handle/2042/13574/A190.pdf?sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
