| Paper: | SS-6.5 | ||
| Session: | Convolutive Blind Source Separation for Speech and Audio Signals | ||
| Time: | Wednesday, May 19, 16:50 - 17:10 | ||
| Presentation: | Special Session Lecture | ||
| Topic: | Special Sessions: Convolutive Blind Source Separation for Speech and Audio Signals | ||
| Title: | BLIND SOURCE SEPARATION IN MOBILE ENVIRONMENTS USING A PRIORI KNOWLEDGE | ||
| Authors: | Erik Visser; SoftMax, Inc. | ||
| Te-Won Lee; SoftMax, Inc. | |||
| Abstract: | A speech enhancement scheme including blind source separationand background denoising based on minimum statisticsis studied in mobile environments. To accommodatethe dependence of the separated output signals on the spatialproperties of the recorded source signals, these blindsignal processing steps are complemented by an adaptiveseparated output channel selection stage using prior knowledgeabout the desired speaker speech content. The resultingscheme performance is illustrated by speech recognitionexperiments on real recordings corrupted by various noisesources and shown to outperform conventional beamformingand single channel denoising techniques as well as anequivalent scheme with fixed output channel selection. | ||
| Back | |||
Home -||-
Organizing Committee -||-
Technical Committee -||-
Technical Program -||-
Plenaries
Paper Submission -||-
Special Sessions -||-
ITT -||-
Paper Review -||-
Exhibits -||-
Tutorials
Information -||-
Registration -||-
Travel Insurance -||-
Housing -||-
Workshops