Sign inSign up
MASK: Robust Local Features for Audio Fingerprinting thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

MASK: Robust Local Features for Audio Fingerprinting

Published on Sep 18, 20122882 Views

Xavier Anguera Miro

This paper presents a novel local audio fingerprint called MASK (Masked Audio Spectral Keypoints) that can effectively encode the acoustic information existent in audio documents and discriminate b

ICME 2012 - Melbourne

Related categories

Feature SelectionComputer Science

Chapter list

MASK: Robust Local Features for Audio Fingerprinting00:00
Outline00:14
What is audio fingerprinting?00:31
What makes a good audio fingerprint?02:37
MASK == Masked Audio Spectral Keypoints03:19
Considered prior art04:53
General Structure06:47
Time-to-Spectral Transformation07:10
Selection of salient spectral points08:00
Spectral masking around salient points09:25
Current MASK regions10:47
Frequency Bands / Time Frames11:44
Fingerprint encoding12:20
Indexing and retrieval13:52
Experimental section14:25
Database14:29
Metric & Baseline15:09
Comparison of Minimum NDCR Scores15:43
Comparison of Actual NDCR Scores16:59
Comparison per transformation17:31
Scores histogram18:03
Conclusions19:11