Bag-of-Audio-Words Approach for Multimedia Event Classification

Abstract

With the popularity of online multimedia videos, there has been much interest in recent years in acoustic event detection and classification for the improvement of online video search. The audio component of a video has the potential to contribute significantly to multimedia event classification. Recent research in audio document classification has drawn parallels to text and image document retrieval by employing what is referred to as the bag-of-audio words (BoAW) method. Compared to supervised approaches where audio concept detectors are trained using annotated data and extracted labels are used as low level features for multimedia event classification. The BoAW approach extracts audio concepts in an unsupervised fashion. Hence this method has the advantage that it can be employed easily for a new set of audio concepts in multimedia videos without going through a laborious annotation effort. In this paper, we explore variations of the BoAW method and present results on NIST 2011 multimedia event detection (MED) dataset.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 13, 2012
Accession Number
AD1038225

Entities

People

  • Murat Akbacak
  • Stephanie Pancoast

Organizations

  • SRI International

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Classification
  • Clustering
  • Detection
  • Detectors
  • Electrical Engineering
  • Event Detection
  • False Alarms
  • Histograms
  • Image Classification
  • Kernel Functions
  • Machine Learning
  • Multimedia
  • Power Spectra
  • Spectra
  • Supervised Machine Learning
  • Warning Systems

Fields of Study

  • Computer science

Readers

  • Computer Networking
  • Neural Network Machine Learning.