Audio fingerprint extraction method using multi-level quantization scheme

Abstract

In this thesis, we propose a new audio fingerprint extraction method, based on Philips music retrieval algorithm, which uses energy difference of neighboring frequency bands and their probabilistic characteristics. Since Philips method uses too many filter-banks in limited frequency band, it may cause audio fingerprints to be highly sensitive to additive noises and to have too high correlation between neighboring bands. The proposed method improves robustness to noises by reducing the number of filter-bank bands while it maintains the discriminative power by representing the energy difference of bands with 2 bits where the quantization levels are determined by probabilistic characteristics. The correlation among 4 different levels in 2 bits is utilized not only in effective similarity measurement, but also in efficient reduction of searching area. The proposed method is evaluated in various noises such as channel noise, environmental noises (street, department, car, office, restaurant) and combined noises. The experiment results show that the proposed method not only gives better performance for environmental noises and highly degraded combined noises, but also takes less time in identifying the music than the Philips method.