Content Based Copy Detection with Coarse Audio-Visual Fingerprints

Abstract

Content Based Copy Detection (CBCD) emerges as a viable choice against active detection methodology ofwatermarking. The very first reason is that the media already under circulation cannot be marked and secondly, CBCD inherently can endure various severe attacks, which watermarking cannot. Although in general, media content is handled independently as visual and audio in this work both information sources are utilized in a unified framework, in which coarse representation of fundamental features are employed. From the copy detection perspective, number of attacks on audio content is limited with respect to visual case. Therefore audio, if present, is an indispensable part of a robust video copy detection system. In this study, the validity of this statement is presented through various experiments on a large data set.

Crim's content-based audio copy detection system for TRECVID 2009

“…The best matching segment is the segment with the highest count. This is similar to the scoring used in [6].…”

“…The first fingerprint that we call energy-difference fingerprint corresponds to the audio fingerprint used in music search and other copy detection tasks [6] [1]. This energy-difference fingerprint that we have used is similar to that used in [6]. It has 15 bits/frame and is extracted as follows:…”

“…We tried two different audio fingerprints. One fingerprinting method is based on energy differences in consecutive sub-bands [6] [1], and results in a very fast search giving good results. The other fingerprints are based on classification of each frame of the test to the nearest frame of the query 1 .…”

“…If we reduce the feature vector from 32 bits to 15 bits, then it occurs frequently enough that we can rely on the matching frame counts as a similarity measure between two segments. For this reason, we experimented with the energy difference parameter used in audio copy detection [6]. We found that this parameter is very robust to various query transformations in the 2008 TRECVID copy detection evaluation [7] and leads to very fast computing.…”

GPU implementation of an audio fingerprints similarity search algorithm

“…The energy difference technique, where a binary fingerprint encodes the energy differences along the frequency and the time axes, figures among the fastest CBCD systems [1,2,3,4]. In [5], regions around selected points from the maxima in the Mel-filtered spectrum are encoded to generate binary fingerprints.…”