Temporal watermarking invertible only through source separation
Gregg Berman, Ismail I. Jouny · IEEE International Conference on Acoustics Speech and Signal Processing · 2002
This paper presents an application of blind signal separation to the field of digital signatures and watermarking of audio signals. The second-order blind identification (SOBI) algorithm presented by Belouchrani et al. can be used to retrieve a watermark or to identify a signature. [1]. Multiple source signals (one of which acts as a digital signature) are arbitrarily mixed with grossly different proportions and transmitted. Only the linear mixtures are observed and no information is known about the prior mixing. This process allows for digital verification of the original message, as well as the ability to transmit hidden, buried messages.