SPREAD SPECTRUMWATERMARKING ROBUST TO SILK VOCODER USING SUB-OPTIMAL DETECTORS

SPREAD SPECTRUMWATERMARKING ROBUST TO SILK VOCODER USING SUB-OPTIMAL DETECTORS

Robust data hiding in speech signals through real-time voice encoders is an interesting but difficult method for watermark- ing. Spread spectrum watermarking is one of the most ro- bust approaches that is implemented in additive and multi- plicative manners. In this paper, a robust detection method is proposed for spread spectrum watermarking in Skype au- dio codec SILK that reduces the error probability through the definition of vectors perpendicular to the key and correlation estimation of the host signal. This method is presented for the additive watermarking and only requires the second or- der statistical moment of the host signal which can be esti- mated at the receiver side. Next, the multiplicative method is turned into an additive one being passed through a non-linear function, and undergoes the proposed method. Moreover, the Skype audio channel effect is considered by SILK output sig- nal statistics being embedded in the detection function.