EUROSPEECH 2003 - INTERSPEECH 2003
Parameter-embedded watermarking of speech signals is effected through slight perturbations of parametric models of some deeply-integrated dynamics of the signal. One of the objectives of the present research is to develop, within the parameter-embedding framework, quantifiable measures of fidelity of the stegosignal and of robustness of the watermark to attack. This paper advances previous developments on parameter-embedded watermarking by introducing a specific technique for watermark selection subject to a fidelity constraint. New results in set-theoretic filtering are used to obtain sets of allowable parameter perturbations (i.e., watermarks) subject to an l_(infinity) constraint on the error between the watermarked and original material. With respect to previous trial-and-error perturbation methods, the set-based parameter perturbation is not only quantified and systematic, it is found to be more robust, and to have a higher threshold of perceptibility with perturbation energy. After a brief review of the general parameter-embedding strategy, the new algorithm for set-theoretic watermark selection is presented. Experiments with real speech data are used to assess robustness and other performance properties. This work is being undertaken in support of the development of the National Gallery of the Spoken Word, a project of the Digital Libraries II Initiative.
Bibliographic reference. Gurijala, A.R. / Deller Jr., J.R. (2003): "Speech watermarking by parametric embedding with an l_(infinity) fidelity criterion", In EUROSPEECH-2003, 2933-2936.