Commercial identification using audio fingerprinting

Abstract

In our study, it is aimed to develop a system using an audio fingerprinting technique, which is used on through their smart phones to identify the commercials played on TV or radio. Thus, users can reach the information about the commercials faster than regular text searching. To achieve this goal, firstly an audio fingerprinting algorithm is developed to find the peaks in the spectrogram and generate the corresponding hashes. In addition to this, three parameters, 1) hash decrease coefficient, 2) matching rate and 3) power rate are introduced, their optimal values are presented and the proposed system is tested with music and commercial databases under various degradations. It is confirmed that hash count and matching duration are decreased, and recall and precision are improved.