An Information-Theoretic Metric of Fingerprint Effectiveness

TitleAn Information-Theoretic Metric of Fingerprint Effectiveness
Publication TypeConference Paper
Year of Publication2015
AuthorsTsai, T.. J., Friedland G., & Anguera X.
Other Numbers3812
Abstract

Audio fingerprinting refers to the process of extracting a robust,compact representation of audio which can be used to uniquely identifyan audio segment. Works in the audio fingerprinting literaturegenerally report results using system-level metrics. Because thesesystems are usually very complex, the overall system-level performancedepends on many different factors. So, while these metricsare useful in understanding how well the entire system performs,they are not very useful in knowing how good or bad the fingerprintdesign is. In this work, we propose a metric of fingerprint effectivenessthat decouples the effect of other system components such asthe search mechanism or the nature of the database. The metric issimple, easy to compute, and has a clear interpretation from an informationtheory perspective. We demonstrate that the metric correlatesdirectly with system-level metrics in assessing fingerprint effectiveness,and we show how it can be used in practice to diagnose theweaknesses in a fingerprint design.

URLhttps://www.icsi.berkeley.edu/pubs/speech/informationtheoretic15.pdf
Bibliographic Notes

Proceedings of the 40th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2015), Brisbane, Australia

Abbreviated Authors

TJ Tsai, G. Friedland, and X. Anguera

ICSI Research Group

Speech

ICSI Publication Type

Article in conference proceedings