Hybrid SVD-based audio watermarking scheme

The issues of audio copyright protection have become very important in recent year. In this paper, a new blind SVD (Singular Value Decomposition) and quantization based audio watermarking scheme in the DCT transform domain is proposed. The watermark information is meaningful binary image permuted by the Piecewise Affine Markov Map (PWAM) before inserted into the host audio signal. The block-based method is applied on the DCT coefficients with the modification of the largest singular values after the SVD transform. The watermark can be extracted without the original signal. Besides, the audio watermarking scheme is adaptive as the quantization step and block length can be changed adaptively without downgrade the signal. The experimental results demonstrate that the proposed scheme is very robust to filtering, extra noise, resampling, requantization, filtering, and Mp3 compression attacks. Besides, the meaningful watermark can still be extracted for protecting rightful ownership after attacks.