| Presentation | 2012-12-21 Reduction of cross spectrum for feature-domain sound source separation Atsushi ANDO, Kenta NIWA, Norihide KITAOKA, Kazuya TAKEDA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Speech source separation is utilized for recognition of simultaneous speech. Conventional source separation methods, especially blind source separation, have a huge computational cost because they require iterative learning steps to estimate separation filters. We therefore try to separate sounds in the feature domain, and the features are then used as inputs for speech recognition, in order to reduce the number of estimated separation filters. For this purpose, linearity between sources and recorded signals is needed in the domain. In this paper, we propose a cross spectrum reduction method between sources to approximate linearity. We prove that taking the average of the power spectra over multiple microphones can reduce the cross spectrum. Experimental results showed that the proposed method could reduce the cross spectrum, and that cepstrum distortions of separated signals were also improved. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | source separation / speech recognition / filterbank outputs / cross spectrum |
| Paper # | SP2012-93 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2012/12/13(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Speech (SP) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Reduction of cross spectrum for feature-domain sound source separation |
| Sub Title (in English) | |
| Keyword(1) | source separation |
| Keyword(2) | speech recognition |
| Keyword(3) | filterbank outputs |
| Keyword(4) | cross spectrum |
| 1st Author's Name | Atsushi ANDO |
| 1st Author's Affiliation | Graduate School of Information Science Nagoya University() |
| 2nd Author's Name | Kenta NIWA |
| 2nd Author's Affiliation | Nippon Telegraph and Telephone Corporation/NTT Media Intelligence Laboratories |
| 3rd Author's Name | Norihide KITAOKA |
| 3rd Author's Affiliation | Graduate School of Information Science Nagoya University |
| 4th Author's Name | Kazuya TAKEDA |
| 4th Author's Affiliation | Graduate School of Information Science Nagoya University |
| Date | 2012-12-21 |
| Paper # | SP2012-93 |
| Volume (vol) | vol.112 |
| Number (no) | 369 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |