Acoustic Event Detection and Localization with Regression Forests

Huy Phan, Marco Maaß, Radoslaw Mazur, Alfred Mertins

Abstract

This paper proposes an approach for the efficient automatic jointdetection and localization of single-channel acoustic events us-ing random forest regression. The audio signals are decom-posed into multiple densely overlappingsuperframesannotatedwith event class labels and their displacements to the temporalstarting and ending points of the events. Using the displacementinformation, a multivariate random forest regression model islearned for each event category to map each superframe to con-tinuous estimates of onset and offset locations of the events. Inaddition, two classifiers are trained using random forest clas-sification to classify superframes of background and differentevent categories. On testing, based on the detection of category-specific superframes using the classifiers, the learned regressorprovides the estimates of onset and offset locations in time ofthe corresponding event. While posing event detection and lo-calization as a regression problem is novel, the quantitative eval-uation on ITC-Irst database of highly variable acoustic eventsshows the efficiency and potential of the proposed approach.
OriginalspracheEnglisch
TitelProc. 15th Annual Conference of the International Speech Communication Association (INTERSPEECH 2014)
Seitenumfang5
ErscheinungsortSingapore
Herausgeber (Verlag)International Speech and Communication Association (ISCA)
Erscheinungsdatum01.09.2014
Seiten2524-2528
PublikationsstatusVeröffentlicht - 01.09.2014
Veranstaltung15th Annual Conference of the International Speech Communication Association: Celebrating the Diversity of Spoken Languages - Max Atria at Singapore Expo Singapore, Singapore, Singapur
Dauer: 14.09.201418.09.2014
Konferenznummer: 108771

Fingerprint

Untersuchen Sie die Forschungsthemen von „Acoustic Event Detection and Localization with Regression Forests“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren