Séminaire de Ms Duong Hien Thanh, doctorante de l'Institut MICA - Date : jeudi 2 juin 2016, 14h00 - Lieu : Salle séminaire, 9ème étage, Institut MICA, Hanoi University of Science and Technology

Intervenant :
Ms Duong Hien Thanh, doctorante de l'Institut MICA

Date : jeudi 2 juin, 14h00
Lieu : salle "seminar room", 9ème étage, Bâtiment B1, Institut MICA, Hanoi University of Science and Technology
Langue : le séminaire sera présenté en anglais

Résumé/abstract:
Acoustic event detection is an important part of the audio analysis system. One of important task of the audio analysis system is content analysis of audio extracting information from an audio recording, to represent its content in terms of sound events taking place. This can be applied in healthcare, such as monitoring of patients or elderly people; security surveillance; detecting violence related events; wildlife monitoring in ecology, etc. To detect acoustic events, we need to have good training data (i.e. audio signals with label of the acoustic event). Thus the goal of my research now is how to automatically extracting unusual sound segments from a long recording in order to help the annotation process easier and quicker. In this presentation, I apply nonnegative matrix factorization (NMF) with one spectral basis to extract acoustic events from long recording, which is recorded outside, includes background noise, environmental noise and other acoustic events (vehicle sound, bird sound, walking sound, speech,...)