Phase discontinuity detection as a means of detecting tampering with speech recordings
Jerome Helffrich, John D. Harrison · The Journal of the Acoustical Society of America · 2012
In the forensic analysis of speech recordings, it is frequently necessary to validate the authenticity of the recording—for example, to determine if it has been altered for the purpose of changing the meaning of what was spoken. One method for doing this is to exploit an incidental steady tonal content in the recording, and look for sudden, discontinuous changes in the phase of that tone. We report on the development of an algorithm based on this idea, to be used for detecting edits in digital or analog speech recordings. Our development process included generating a corpus from the TIMIT speech database, designing an algorithm to exploit possible phase discontinuities in the recorded signals at edit sites, and testing the algorithm against the corpus with both edited and unedited copies of the sentence material. A statistical analysis of the performance of the algorithm on sentences contaminated by noise and compressed by various digital compression schemes will be given.