Development of new frequency techniques for audio signal watermarking

Abstract

In this thesis, we present three contributions in the field of digital audio watermarking in the frequency domain. In the first contribution, we propose a parametric quantization index modulation (QIM) for which we find the optimal values of the parameters using the Lagrange multipliers method. Moreover, we present an approach for selecting the embedding positions in the discrete cosine transform (DCT) domain that gives the immunity against low- and high-pass filtering attacks. Furthermore, a fast algorithm for the proposed technique is developed. In the second contribution, we introduce a QIM-based technique that unifies blind and semi-blind audio watermarking in the discrete wavelet transform (DWT) domain. We also propose an expression that gives a quantization step, which is invariant under additive white Gaussian noise and renders the minimum distance decoder of QIM invariant under amplitude scaling of the host signal. Moreover, to avoid the usage of an extra channel, we propose an efficient side-information recovery procedure for the semi-blind mode of operation of the proposed technique. In the third contribution, we propose a robust-hashing method by using QIM’s minimum distance decoder in the DWT domain, then we propose a joint hashing-watermarking technique for audio fingerprinting to enhance the perceptual quality and robustness while reducing the storage requirement compared with other audio watermarking techniques. Finally, we conduct various experiments to validate the correctness of the theoretical expressions derived in this thesis and also to evaluate the proposed techniques and compare them with existing relevant techniques