Speaker: Nikolaos Vasiloglou
Title: A fingerprint technique for extracting  midi sounds from an auditory scene
Date: September 19, 2003
Time: 3:00 pm
Location: GCATT Room 325
Abstract: There are several ways to approach the problem of the
Auditory Scene Analysis  (ASA). Human perception and cortex processing have inspired several researchers to develop algorithms that can partially solve the problem under very tight assumptions. Psychoacoustic experiments have shown that the segregation of sound sources is boosted if the sounds are already known. In this presentation we approach the problem of separating midi sounds from a rich auditory scene. The algorithm is based on the fact that there is prior knowledge of the midi source. A variation of an audio fingerprint method developed by PHILIPS is used in order to index prerecorded midi sounds in a database.


Biography:


Slides:  sem09_05_2003_Nikolaos_Vasiloglou.swf