Stochastic model based audio watermark and whitening filter for improved detection

Digital watermarking is a means of copyright protection by embedding an imperceptible signal into the multimedia data. Nearly all watermarking schemes developed basically employ a spread spectrum technique, and the extraction of embedded watermark depends on the correlation method. In this paper, we propose a stochastic model based public audio watermark. Based on the stochastic model and detection theory, the watermark shaping effect on detection is well explained. We also provide the theoretical whitening filter SNR gain and show that the detection performance can be significantly improved with the whitening filter. As regards robust detection with the whitening filter, we propose a robust detection parameter and demonstrate the reliability of the proposed estimated whitening filter to various kinds of audio degradation and distortions, e.g., MPEG compression, filtering, requantization, resampling, etc