Abstract:

To detect resampling in audio tampering, a new spectral measure, band-partitioning spectral smoothness (BPSS), is proposed to measure the degree of spectral fluctuation. Based on the analysis to spectral characteristics of non-interpolated and interpolated audio signals, a comprehensive strategy to detect audio interpolation and stitching with BPSS is presented. Experiment results show that, by using the proposed method, it can be correctly determined whether the audio signals are interpolated or stitched. Furthermore, the original sampling rates of the interpolated signals can be estimated with high accuracy. Low computational complexity enables this method to be useful in real time audio tampering detection.

Key words: audio tampering detection, interpolation, stitching, band-partitioning spectral smoothness