A Digital Audio Watermarking Scheme Based on Independent Component Analysis and Singular Value Decomposition

A robust audio watermarking scheme based on independent component analysis (ICA) and singular value decomposition (SVD) is presented in this paper. ICA is a novel method of signal processing and data analysis, through which the source signals can be derived from the observed signals even in the case of the source signals and the mixing procedure being unknown, so long as they are independent. SVD is a linear algebra technique used to diagonalizable matrices and it packs most of the signal energy into very few singular values (SVs). In this paper, the original watermark, which is a binary image, is rearranged by Arnold transform before embedded into the host signal. In the embedding procedure, the discrete cosine transform (DCT) coefficients of the host signal and the embedding watermark, which are regarded as two sources are mixed through the ICA instantaneous mixing model. Then, two mixed signals can be obtained. One of them is used to generate the watermarked signal and the other is kept as the secret key. In the extraction procedure, with the watermarked signal and the secret key, the popular ICA algorithm is employed to get the estimation of the embedding watermark. Experimental results show that the proposed method has good robustness under general signal processing