Accelerating Content Based Music Retrieval Using Audio Fingerprinting

Abstract

Audio Fingerprinting is a content-based signature to refer from original audio file. Audio Fingerprinting technique is commonly used to monitor the audio without the need of meta-data. Fingerprinting approaches are Pattern matching, Multimedia Information Retrieval and Cryptography. For browsing and identifying the audio content from large audio data collection set is a major strand for research. The best solution for identifying the audio content is Audio ID System in which the specified audio is identified by passing a short query audio clip from a large set of audio data set. As the main contribution, the system establishes the connection between the hash values which are mainly used for indexing and their ability to survive the presence of noise and distortions. This improves the performance of the audio ID system by using following four ways: First, the fingerprints are generated to increase its reliability. Second, by providing the priority to reliable hash values which are basically use during searching for reference entries, it improves the retrieval speed. Third increase the identification of reliable hash values by enlarging the query fingerprint. Fourth indexing only the reliable hash values, this lowers the server side storage requirements.