Digital audio recording forgeries are often created using simple editing techniques such as ‘butt splicing’. Butt splicing may leave discontinuities in the audio waveform that may or may not be audible. Detection of butt-spliced edits will be discussed and a new method to detect edits made in this manner will be proposed. The process operates in the time domain and is based on high pass filtering the audio data and modelling a discontinuity at higher frequencies where the ratio of discontinuity energy to acoustic signal energy level is improved. The model is then used as a template to search for potential edits in the filtered audio signal. The technique is optimised for data that has not been perceptually encoded post editing and is capable of detecting discontinuity points within a recording that are not discernable by auditory analysis.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.