Robust Audio Watermark Detection Based on Dempster-Shafer Theory of Evidence

摘要

Watermark detection algorithm plays an important role in enhancing the robustness of watermark. This paper proposes one watermark detection technique based on Dempster-Shafer (D-S) theory. D-S theory can be simply viewed as an extension to classical Bayesian Theory in statistical inference. The proposed audio watermark detection method uses D-S theory to model uncertainty when the watermarked audio has been attacked, by using D-S's rule of combination, correlation results from several frames are combined to form a more reliable decision. Results from simulation experiment using spread spectrum techniques and audio masking validate the proposed detection algorithm.