Sub-fingerprint masking for a robust audio fingerprinting system in a real-noise environment for portable consumer devices

Abstract

The robustness of audio fingerprinting system in a noisy environment is a principal challenge in the area of content-based music retrieval, especially for use in portable consumer devices. Our new audio fingerprint method using sub-fingerprint masking based on the predominant pitch extraction dramatically increases the accuracy of the audio fingerprinting system in a noisy environment, while requiring much less computing power for matching, compared to the expanded hash table lookup method, where the searching complexity increases by the factor of 33 times the degree of expansion.