Subjective and objective quality assessment of re-sampling audio

Audio tampering is often accompanied with resampling especially in audio splicing, re-compression, time-scale and pitch-scale modification etc. We assess the quality of re-sampling audio by two measures, in which Extended Bark Spectral Distortion model (EBSD) is adopted as an objective measure and Mean Opinion Score (MOS) is selected as a subjective measure. Experiments show that EBSD measure has a high correlation with MOS score, which reflects that quality of re-sampling audio is nearly the same with original audio when re-sampling factor is equal or greater than 0.5, but poor when factor is less than 0.5. Therefore, we can distinguish original audio from resampling audio by quality when the factor is small, that can be effectively used as auxiliary information in audio forensics.