Watermarking of audio signals using independent component analysis

A novel approach to watermarking of audio signals using Independent Component Analysis (ICA) is proposed. It exploits the statistical independence of components obtained by practical ICA algorithms to provide a robust watermarking scheme with high information rate and low distortion. Numerical simulations have been performed on audio signals, showing good robustness of the watermark against common attacks with unnoticeable distortion, even for high information rates. An important aspect of the method is its domain independence: it can be used to hide information in other types of data, with minor technical adaptations.