Abstract:

Insertion and deletion are two commonly-used audio forgery operations,and the current Electric NetworkFrequency( ENF) -based audio forgery detection algorithms cannot accurately determine the location of insertion anddeletion operations.In order to solve this problem,a blind forgery detection algorithm based on dual judgmentmechanisms is proposed in this paper.In the investigation,the first mechanism utilizes the information from aMOCC ( max offset for cross correlation) curve to locate the forgery while the second mechanism estimates forgeryregions according to the slope curve which corresponds to the MOCC curve.The combined use of the two mechanismscan obtain a higher forgery localization precision.For simplicity of implementation,a method is put forward to calculatethe MOCC without an extra reference ENF signal.As compared with two current representative methods,the proposeddetection method can locate the deletion and insertion operations more accurately.Besides,it has reasonablerobustness when applied to such common audio operations as re-sampling,compression and noise addition.

Key words: audio forgery detection, electric network frequency, max offset for cross correlation, slope, dual judgment mechanisms