The phase of the electric network signal (ENS) contains valuable information regarding a potential editing in the analyzed audio recording. When the fundamental component of the ENS is deliberately erased, an analysis of its harmonics, assuming they are present in the signal, can be performed instead. In this context we propose to detect an audio editing by tracking the phase discontinuities of the ENS harmonics (ENSH) not over the entire signal, but only over the silence segments from a speech recording. A direct consequence of our approach is the separation of the ENSH components from dominant speech components, which in turn allows a better detection and estimation of the ENSH parameters. We complete the algorithm with a simple energy based criterion for the ENSH detection, which in conjunction with a high accuracy spectral analysis method based on the instantaneous frequency attractors, leads to superior results in terms of audio editing detection rate. Compared with similar methods, the proposed algorithm is less time-consuming and provides good detection rates even for speech signals rich in low frequency content (e.g. male voices).

Keywords: Harmonics of the electric network signal, Instantaneous frequency attractors