ROBUSTNESS OF AUDIO FINGERPRINTING SYSTEMS FOR CONNECTED AUDIO APPLICATIONS

摘要

Audio fingerprinting allows the identification of audio content regardless of the audio format and especially without need of additional meta-data. A Well-known applications of audio fingerprinting are connected audio. Music identification using a mobile phone is one these examples. The robustness of the audio fingerprinting to the distortions induced by the different components of mobile network is important to guarantee good performance of the audio identification process. This paper focus on the distortions induced by speech codecs encountered by music signal during a radio-mobile transmission. The robustness of two selected fingerprinting systems is analyzed with respect to the real GSM speech codec. In particular, it will be shown through the analysis that rhythmical parameters seems to be more robust than timbrical ones.