Dominant feature vectors based audio similarity measure

Authors:
Jing Gu;Lie Lu;Rui Cai;Hong-Jiang Zhang;Jian Yang
Affiliations:
Tsinghua Univ., Dept. of Electronic Engineering, Beijing, China;Microsoft Research Asia, Beijing, China;Dept. of Computer Science and Technology, Tsinghua Univ., Beijing, China;Microsoft Research Asia, Beijing, China;Tsinghua Univ., Dept. of Electronic Engineering, Beijing, China
Venue:
PCM'04 Proceedings of the 5th Pacific Rim Conference on Advances in Multimedia Information Processing - Volume Part II
Year:
2004

Citing 3
Cited 5

Distance measures for signal processing and pattern recognition

Signal Processing
Content-Based Classification, Search, and Retrieval of Audio

IEEE MultiMedia
Using structure patterns of temporal and spectral feature in audio similarity measure

MULTIMEDIA '03 Proceedings of the eleventh ACM international conference on Multimedia

Audio similarity measure by graph modeling and matching

MULTIMEDIA '06 Proceedings of the 14th annual ACM international conference on Multimedia
Combining multimodal preferences for multimedia information retrieval

Proceedings of the international workshop on Workshop on multimedia information retrieval
Repeating pattern discovery from audio stream

ICMLC'05 Proceedings of the 4th international conference on Advances in Machine Learning and Cybernetics
Using earth mover's distance for audio clip retrieval

PCM'06 Proceedings of the 7th Pacific Rim conference on Advances in Multimedia Information Processing
Learning user queries in multimodal dissimilarity spaces

AMR'05 Proceedings of the Third international conference on Adaptive Multimedia Retrieval: user, context, and feedback

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an approach to extracting dominant feature vectors from an individual audio clip and then proposes a new similarity measure based on the dominant feature vectors. Instead of using the mean and standard deviation of frame features in most conventional methods, the most salient characteristics of an audio clip are represented in the form of several dominant feature vectors. These dominant feature vectors give a better description of the fundamental properties of an audio clip, especially when frame features change a lot along the time line. Evaluations on a content-based audio retrieval system indicate an obvious improvement after using the proposed similarity measure, compared with some other conventional methods.