TV Advertisement Search Based on Audio Peak-Pair Hashing in Real Environments

This paper proposes a robust TV advertisement search based on audio fingerprinting in real environments. To obtain prominent audio peak pairs against different types of distortions, an adaptive thresholding method based on a median filter and peak-picking update is applied. Using the prominent audio peak-pair hashing, the proposed audio fingerprinting algorithm improves the robustness of the TV advertisement search against noise, pitch-shifting, and time-stretching, and delivers high identification accuracy in spite of short queries. The experimental results show that the proposed method has respectable results compared to other methods.