Spread-transform dither modulation for audio watermarking and its improvements with LDPC codes

Quantization watermarking methods are well known for its high payload, however its robustness is limited to the quantization step size. In this paper, we propose several audio watermarking techniques based on quantization watermarking and LDPC codes which achieve good robustness. In specific, using an half-rate Margulis LDPC code we can obtain a benefit of 5 dB in robustness against additive white Gaussian noise. In distortion-compensation schemes with LDPC codes is possible to obtain better robustness with scale parameter alpha,α < α _opt* lower than the theoretically optimal. Our algorithms keep a payload higher than 20 bps and the watermarked audio quality achieves a subjective grade higher of 4-0 of 5.0 possible.