|
Human and Machine Hearing: Extracting Features from Sound| old_uid | 8806 |
|---|
| title | Human and Machine Hearing: Extracting Features from Sound |
|---|
| start_date | 2010/06/01 |
|---|
| schedule | 16h30-17h30 |
|---|
| online | no |
|---|
| location_info | 1er étage, salle Paul LApie |
|---|
| summary | We develop a teachable four-part system structure for machine hearing applications. An auditory periphery model, based on the pole-zero filter cascade (PZFC) connects psychophysical and physiological models of nonlinear auditory filtering with an efficient digital algorithm. An auditory image stage creates one or more representations of the sort that the brainstem and midbrain send to auditory cortex. A sparse feature extractor models abstractly the action of cortical feature detection cells. Finally, a trainable machine-learning layer extracts the decisions needed to suit an application. Applications to speech, music, security and surveillance systems, personal sound diaries, smart house, etc., are anticipated. A planned book and course will be previewed. |
|---|
| responsibles | Pressnitzer |
|---|
| |
|