A novel adaptive watermark embedding approach for enhancing security of biometric voice/speech systems

In this paper, we propose a novel watermarking method in order to enhance the security in biometric voice/speech transmission systems basing on sub-sampling, discrete cosine transform (DCT) and adaptive watermark embedding. To improve imperceptibility, we use sub-sampling and adaptive embedding in DCT high energy coefficients. We employ a significant watermark represented in a biometric unique fingerprint. Introducing bits, only in high energy fraction, provides us with further agreeing element depicted by decreasing running time, both in embedding and extraction processes, which can help minimizing hardware consumption. Achieved results reveal the stability and flexibility of our proposed scheme and confirm its robustness against additive noise. In addition, we enhanced our previous published approaches and the expected limitations of our proposed model will appear in the hardware implementation where the attacks will be more considered. Another limitation will be the number of quantization bits used for high energy DCT coefficients. All these may affect the speaker secured identification and verification system.