|
Robot audition from the viewpoint of computational auditory scene analysis| old_uid | 1505 |
|---|
| title | Robot audition from the viewpoint of computational auditory scene analysis |
|---|
| start_date | 2006/07/04 |
|---|
| schedule | 13h30 |
|---|
| online | no |
|---|
| location_info | bât. C, 1e étage, amphi C |
|---|
| summary | In this talk, I'd like to overview my research towards listening to several things at once. First, I'll present three issues of robot audition, sound source localization, separation, and recognition of separated speech signals. Second, I'll present a real-time speaker tracking system on a robot that can track multiple speakers by hierarchical integration of audition and vision by using two microphones.
Third, I'll present three kinds of sound source separation systems, active direction-pass filter with a pair of microphones, geometric source separation with eight microphones, and independent component analysis with a pair of microphones. The separated sound is recognized by missing-feature theory based ASR. Missing feature masks are generated by estimating leak energy between separated sounds. My talk will be accompanied by several demonstration vidios. |
|---|
| responsibles | Lécuyer |
|---|
| |
|