Audio authenticity based on the discontinuity of ENF higher harmonics

This work deals with the use of electrical network frequency (ENF) higher harmonics to assess audio authenticity. It is assumed that ENF has corrupted the signal under analysis and also that, due to some non-linearity inherent to the sound recording process, higher harmonics are present. After down-sampling the audio signal and band-filtering it around a higher harmonic frequency, we estimate the phase of the resulting signal; the result provides a visual aid to check for possible audio tampering indicated by abrupt phase changes. As in a previous work based on the nominal ENF, we here use a feature extracted from the signal estimated phase to perform an automatic audio authenticity test. The result, although not as accurate as when the nominal ENF is used, provides a useful hint to audio forensic analyst whenever, for some reason, the nominal ENF has been removed from the signal.