A Novel Recognition Method for Compressed-domain Audio
Ming Jian-chen · Science Technology and Engineering · 2014
As the MPEG compression format becomes the main form of audio file storage and transmission,audio fingerprint identification technology for compressed-domain audio appears to be more practical significance. A novel compressed-domain audio fingerprint algorithm is proposed,which can be applied to compressed-domain audio fingerprint and uncompressed-domain audio fingerprint retrieved in the same fingerprint database effectively.Experiment chooses the most typical audio compression format——MP3 audio to test the property of the algorithm.Experimental results show that the proposed algorithm has smaller fingerprint size,better fingerprint distinction and higher correct recognition rate. Moreover,it has strong robustness to a variety of common time-frequency domain distortions.