Abstract: Audio tampering is often accompanied with re-sampling especially in audio plicing, re-compression, time-scale and pitch-scale modification, quality faking and other perations. Therefore, re-sampling detection is an important means in audio forensics. his paper proposes a new method to detect audio re-sampling. It is theoretically shown hat, if an original audio signal is re-sampled, significant peaks can be found in second-order erivative of the spectrum, and the peak position is related to re-sampling factor. This ethod is applicable to common nonlinear interpolation. Detection performance influenced by audio duration or content or format is discussed. Finally, an automatic detection scheme is described, and experimental results show that the algorithm can be used to detect resampled audio effectively.

Key words: linear interpolation, nonlinear interpolation, re-sampling detection, second-order derivative, audio forensics