An efficient audio watermark by autocorrelation methods

Abstract

Some audio watermark applications, specifically the second screen, are subject to harsh environment, such as high attenuation and interference during the transmission of the stego signal from speaker of the primary screen to microphone of the second screen. Autocorrelation domain embedded audio watermarking systems are a promising algorithm since the watermark signal is highly correlated to the host signal causing an inherently temporal masking effect. At the same time the algorithm promises blind watermark data extraction since the watermark data are hidden in the relationship between the stego signal and its delayed version. Only if the stego key remains secret, the algorithm satisfies the three criteria of an audio watermark algorithm: robustness, imperceptibility, and capacity. In this paper, we discuss several derivations of the algorithm and present the performances of the algorithm in terms of bit detection rate for some parameter arrangements. Discussion about the robustness of the algorithm performances under attacks is also presented.