Copy-paste Tampering Detection Algorithm for Audio Signals Based on Hilbert-Huang Transform

Copy-paste tampering is a common method for audio tampering, which is often difficult to detect when the tampered audio segment comes from the same audio accompanied by post-processing operations. For audio signal copy-paste tampering with post-processing, this paper proposes a method for detecting and locating audio signal copy-paste tampering based on the Hilbert-Huang transform. Firstly, the pitch tracking technique is used to distinguish the voiced and unvoiced segments of the audio signal, and the Hilbert-Huang transform is applied to each voiced segment separately to extract the spectrogram features, and then dynamic time warping (DTW) is used to calculate the similarity of each spectrogram feature. When a certain threshold is met, the audio copy-paste tampering is detected by comparing the DTW values of the Hilbert spectrogram features and locating the location of the copy-paste tampering according to the index of the DTW value corresponding to the audible segment. The experimental results show that the method has good robustness to several common audio signal post-processing operations after copy-paste tampering. The proposed method has been demonstrated through extensive experiments to have higher detection accuracy, recall, and robustness compared with Kasi's proposed Mel-spectrogram-based audio signal copy-paste tampering detection method.