Extracting Quality Parameters for Compressed Audio from Fingerprints

An audio fingerprint is a compact yet very robust representation of the perceptually relevant parts of audio content. It can be used to identify audio, even when of severely distorted. Audio compression causes small changes in the fingerprint. We aim to exploit these small fingerprint differences due to compression to assess the perceptual quality of the compressed audio file. Analysis shows that for uncorrelated signals the Bit Error Rate (BER) is approximately inversely proportional to the square root of the Signal-to-Noise Ratio (SNR) of the signal. Experiments using real music confirm this relation. Further experiments show how the various local spectral characteristics cause a large variation in the behavior of the fingerprint difference as a function of SNR or the bitrate set for compression.

• Type: Conference
• Year: 2005
• Where: ISMIR
• Authors: Peter Jan O. Doets, Reginald L. Lagendijk
• Keywords: Audio Compression, Audio fingerprint, Information Retrieval, ISMIR 2005, fingerprint Difference