Compact and Robust Audio Fingerprinting for Speedy Music Identification
Abstract— An audio fingerprint, which is a compact content-based digest of an audio signal, is widely used to quickly locate perceptually similar songs in an audio database. For a million -song library, memory imposes a restriction for speedy and correct music identification and thus demands a compact fingerp rinting system. This paper focuses on reducing memory requirement of fingerprint storage while preserving the robustness of fingerprints to common distortions such as compression, noise addition, etc. In this system, a 3- sec audio clip is represented by a 2712-bit fingerprint block. It significantly reduced the storage when compared with the Philips Robust Hashing (PRH), one of the dominant researches of audio fingerprinting, where a 3-sec audio clip was represented by an 8192-bit fingerprint block. Experimental results also showed that the reliability and robustness of the proposed fingerprinting system outperforms the PRH under various distortions, especially linear speed changes and pitch shifting.
Index Terms— audio fingerprint, a million-song library, music identification, Philips Robust Hashing.
Myo Thet Htun, Twe Ta Oo
Faculty of Computer Systems and Technologies, University of Computer Studies, MYANMAR
Cite: Myo Thet Htun, Twe Ta Oo, "Compact and Robust Audio Fingerprinting for Speedy Music Identification," Proceedings of 2019 the 9th International Workshop on Computer Science and Engineering WCSE_2019_SPRING, pp. 48-57, Yangon, Myanmar, February 27-March 1, 2019.