Optimal selective correlation detector for spread-spectrum watermarking systems

This paper proposes a new optimal correlation detector for additive spread spectrum watermarking systems. The proposed detector divides the signal into blocks, on each of which detection is made and each block is again divided into several sub-blocks. It selects in each block the optimal number of sub-blocks with highest watermark to noise ratio (WNR) and performs the correlation detection only on the selected sub-blocks. This corresponds to minimizing the variance of the correlation. Experiments are conducted on audio signals of several different audio genres in the discrete fourier transform (DFT) magnitude domain, and the results show that the proposed detector exhibits strong robustness to various attacks as well as dramatic improvement in performance.