Fingerprint Producing Method and Audio Fingerprinting System Based on Normalized Spectral Subband Centroids

摘要

The present invention provides a normalized spectral sub-band center (Normalized Spectral Subband Centroid; NSSC) and based on fingerprint and the method, an unknown audio signal is given as input to generate, retrieve it from the database that is already built with large audio for audio fingerprinting system is intended to output the information about the input audio signal. The audio fingerprinting system is that by using the fingerprint performance greatly affected. NSSC fingerprint of the present invention is suitable for use in the recognition of the characteristics of the audio signal represented by a well, while a number of distortions that can be applied to the audio signal, such as an MP3 compression, equalization, etc., and very strong (equalization), and building a large database advantageous to have a real-time search. According to the experimental results, the system according to the present invention exhibit improved performance compared to conventional audio fingerprinting system. Real-time filtering service, automatic indexing of large audio database for the present invention is to prevent illegal distribution of audio on the Internet (indexing), and can be applied and broadcast monitoring.