205 An ASF and DCT Based Audio Feature Extraction Algorithm

Audio fingerprinting techniques aim at successfully performing content-based audio identification even when the audio signals are slightly or seriously distorted. And the fingerprint extraction algorithm is a great important part of this system. In this paper, we present a new feature extraction algorithm, which is based on the Discrete Cosine Transform (DCT) and Audio Spectrum Flatness (ASF). The ASF is a feature of the MPEG-7 standard, which are new to the audio feature family and have not been considered as much as other feature types. DCT is applied to the sub-band ASF of each frame for it has a strong energy compaction property and it is extremely close to the Karhunen-Loéve transform. Experimental results show that the proposed algorithm can properly match the right position of the audio. And it is robust to some distortions such as echo addition, MP3 compression, down sample, up sample and equalization. Meanwhile, it also has good granularity property for it only needs 4.2s to identify the audio.