Robust audio copy-move forgery detection on short forged slices using sliding window

Abstract

Audio copy-move forgery detection (CMFD), which aims at finding possible forgeries that are derived from the same audio recording, has been one of the most difficult challenges in blind audio forensics. However, the existing works rarely pay attention to the forgery of short slices within voiced segments, which may result in missed detection. In this work, we present a robust audio CMFD method on the basis of sliding window (SW) and constant Q cepstral coefficients (CQCC). Specifically, we first propose two SW strategies to illustrate how to reveal the possible copy-move forgeries within and between voiced segments. Then, we combine the CQCC feature with the Pearson correlation coefficients to measure the similarity between different short slices. Finally, we evaluate the proposed method, named SW-CQCC, against the state-of-the-art approaches to CMFD on real-world read English and Chinese corpora. Experimental results demonstrate that our SW-CQCC exhibits significantly high effectiveness and robustness in detecting short forged slices within voiced segments, which may be helpful to audio forensic examinations.