数字水印已经成为实现多媒体版权保护的一项重要技术.提出在嵌入鲁棒性水印的同时考虑攻击的行为特征,辅以陷阱策略抵制各种攻击.水印以量化音频DCT系数的方式嵌入,在提取时针对攻击的行为特征以不同方式提取水印.算法嵌入的水印为一幅二值图像,能抵制随机裁剪、AD及DA转换、低通滤波等去同步型攻击,提取的水印可以直接分辨.与cocktail水印方法相比,所提方法的鲁棒性高,提取水印时不需要原始音频.实验结果说明我们的音频水印方法十分有效.

With the development of multimedia technology and computer network, the distribution and transmission of digital media (e.g., image, text, audio and video) become very convenient. But rampant piracy, forgery and tamper is stunting the development of digital era. How to provide available method for copyright protection and information Security has become a hot research area. Digital watermarking as a copyright protection technology is developed in this condition. Digital audio watermarking is especially challenging compared with digital video and image watermarking because the human auditory system (HAS) is far more sensitive than the human visual system (HVS). An audio watermarking algorithm based-on Modulated Complex Lapped Transform (MCLT) is proposed. Using spectrum-warping technique that shifts spectrum coefficients non-linearly, a significant binary image is embedded in audio by modulating MCLT's angle frequency. In order to improve the security, some pretreatments for watermark are carried through before it is embedded to audio, such as repeat coding and interleaver. The watermarking is embedded in audio frequency domain, and embedding capability is high. The experimental results show that the spectrum-warping watermarking algorithm is of low robustness against MP3 compression, resampling and so on. It is a fragile watermarking algorithm.