A multi-channel approach through fusion of audio for detecting video inter-frame forgery

The forgery operation of digital video in the temporal domain is often accompanied by the synchronization of the audio channel operation. In this paper, we proposed a fusion of audio forensics detection methods for video inter-frame forgery. First, the audio channel of the video is extracted, and discrete wavelet packet decomposition and analysis of singularity points of audio signals are used to locate the forged singularity points. Next, features of each frame of the video are extracted with the perceptual hash and used to calculate the similarity between consecutive frames, to locate the forgery position in the video frame sequence. We fused the results of the audio channel and the video frame sequence channel. The QDCT feature is used to further fine detect the suspected forgery location. Our method can position replication source locations for copy-move forgery. Experiments show that our method has higher accuracy and better performance in comparison with similar methods, especially on the delete forgery operation.