IMPROVING AUDIO WATERMARKING PERFORMANCE BY CARRIER MODULATIONS

摘要

In this paper, we propose to evaluate the performance of different modulation methods in audio watermarking system. Unlike the most existing methods, that use spread spectrum modulation, we propose to construct the watermark by using bandpass modulations that are widely used in digital communication, but never introduced to our knowledge in audio watermarking systems. The watermark exploits the perceptual frequency masking of the Human Auditory System (HAS) and adapts the bandpass signal (watermark) parameters in order to make the watermark inaudible. Therefore, the watermark is recovered without any knowledge of the original audio signal. A comparative study between bandpass modulations allows us to make the choice on the best one that ensures a nice compromise between inaudibility constraint and detection reliability of the embedded information. Experimental results exhibit that an expressive compromising, in terms of watermark transparency and high robustness of the watermarked audio can be achieved.