Voice biometric watermarking of 3D models

Digital watermarking provides a solution to piracy and copyright protection of 3D multimedia content by embedding a hidden piece of information in the original content. This hidden piece of information travels with the 3D multimedia in the distribution chain and assists in verifying legitimate ownership and tracing piracy. This paper proposes a voice-based biometric watermark for 3D graphics for the purpose of owner identification, traitor tracing and access control schemes. A voice print is generated using a Gaussian mixture model representation of Mel-frequency cepstral coding coefficients of an individual´s speech. This voice print is inserted as the watermark. The proposed technique generates a semi-fragile watermark, which is tolerant to a designated class of transformations. Experimental results indicate that the biometric watermark is resistant to cropping, low levels of Gaussian noise addition, and is intolerant to mesh smoothing attacks.