An algorithm of detecting audio copy-move forgery based on DCT and SVD

In order to detect whether there is copy-move forgery in an audio file, a new algorithm is proposed in this paper. In this algorithm, some syllables are obtained by segmenting audio signal with voice activity detection (VAD) algorithm. Discrete cosine transform (DCT) is performed to process these syllables to generate DCT coefficients. Then these coefficients corresponding to every syllable are converted to a square matrix. After that, singular value decomposition (SVD) transform is performed for the square matrix to obtain singular eigenvector. Here, the contribution rate is calculated to reduce the dimension of the singular eigenvector. So the amount of data to be processed can be reduced greatly. Finally, the distances among syllables are measured by comparing the distance among the low-dimensional singular vectors to determine whether there is the copy-move relationship between corresponding syllables. Because these singular values have a great stability, the distance between two syllables would have little change after the common signal processing. The low-dimension singular eigenvector requires less computational complexity, so the algorithm costs less running time. The experiment results show that the proposed algorithm is robust against conventional attacks and has better efficiency.