自适应的混合域音频水印新算法

摘要

An adaptive audio watermarking algorithm in the hybrid domain is proposed,the scheme jointly exploiting the discrete wavelet packet transform (DWPT) and psychoacoustic model to perform large-capacity audio watermarking without introducing perceptible distortion.Firstly,each audio frame was decomposed into 26 sub-band signals by DWPT,then we apply DCT to wavelet packet coefficients of each sub-band and calculate the masking threshold.The masking threshold of each sub-band was used to seek suitable segments and positions for watermark embedding.According to the embedding strength obtained from psychoacoustic model,the algorithm adaptively control the audibility of introduced distortion for embedding the watermark.The binary image watermark was embedded into the block middle-frequency and low-frequency DCT coefficients according to quantization index modulation,the extraction was executed blindly.Experimental results show that the proposed algorithm achieves a good trade-off between robustness,imperceptibility and payload,the watermark capacity range from 576.7 bps to 689.5 bps,and the hidden watermark data is robust to additive noise,re-quantization,re-sampling,low-pass filtering,and MP3 compression.%本文结合小波包变换和心理声学模型,提出了一种自适应的混合域音频盲水印算法,在不引入明显听觉失真的前提下,实现了大容量的水印嵌入.算法首先采用小波包变换将分段音频信号分解到26个子带中,然后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水印嵌入段和水印嵌入位置,同时根据掩蔽阈值计算出的水印嵌入强度自适应地控制由水印嵌入引起的听觉偏倚.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.实验结果表明本算法在水印容量、不可感知性和鲁棒性之间达到了很好的平衡,水印容量在576.7 bps到689.5 bps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.