In echo watermarking, effort to improve robustness often conflicts with the requirement of imperceptibility. Theyhave been inherent trade-offs in general audio watermarking techniques. In this paper, we challenge to the developmentof imperceptible but detectable echo kernels being directly embedded into the high-quality audio signal.Mathematical and perceptual characteristics of echo kernels are analyzed in frequency domain. Finally, we canobtain much flat frequency response in perceptually significant bands by combining closely located positive andnegative echoes. The proposed echo makes it possible to improve the robustness of echo watermark withoutbreaking the imperceptibility.
展开▼