自适应的混合域音频水印新算法

本文结合小波包变换和离散余弦变换，提出了一种基于听觉模型的混合域自适应音频盲水印算法，在不引入听觉失真的前提下，实现了自适应的水印嵌入。算法首先对音频信号进行小波包分解，使得分解后的子带更接近人耳临界频带。其次对每个子带的小波包系数进行离散余弦变换，计算出子带掩蔽阈值。根据子带掩蔽阈值自适应的选取噪声敏感度小的音频段作为水印嵌入段，选取功率值低于掩蔽阈值的频域系数作为水印嵌入位置，同时采用噪声掩蔽比调整水印嵌入强度。二值水印图像通过量化索引调制的方法嵌入到音频信号的中低频系数中，提取水印时不需要原始音频载体。本算法在水印容量、不可感知性和鲁棒性之间达到了很好的平衡，水印容量在576.7bps到689.5bps之间，算法对添加噪声、重新量化、重新采样、低通滤波和MP3压缩均具有很好的鲁棒性。 

Adaptive audio watermarking algorithm in the hybrid domain

An adaptive blind digital audio watermarking algorithm based on psychoacoustic model is proposed. The scheme jointly exploiting the discrete wavelet packet transform (DWPT) and the discrete cosine transform (DCT) to perform large-capacity blind audio watermarking without introducing perceptible distortion. Each audio frame was divided into 26 sub-band signals by using DWPT. After that, we apply DCT to wavelet packet coefficients and calculate the masking threashold. The masking threshold of each sub-band was used to seek suitable segments and position for watermark embedding. The binary image watermark was embeded into the block middle-frequency and low-frequency DCT coefficients according to quantization index modulation. The quantization step of DCT coefficient in each sub-band was determined by the sub-band ratio of noise to masking threshold .The extraction was executed blindly. Experimental results show that our scheme achieves a good trade-off between robustness, imperceptibility and payload, the watermark capacity range from 576.7bps to 689.5bps,the hidden watermark data is robust to additive noise, requantization, resampling, low-pass filtering, and MP3 compression.