基于邻域平均的鲁棒音频水印算法

Robust Audio Watermarking Algorithm Based on Neighborhood Averaging Method

针对音频作品的版权保护问题,从时域角度提出一种鲁棒水印算法。使用邻域平均的方法,算法选择两个合适的正整数,每次移动一个采样点,分别计算连续采样点的均值,所得两个均值序列构成两条均值线,这两条均值线在时域上会出现交叉。然后用量化索引调制(QIM)方法,把水印信息嵌入在均值线的交叉点上。实验表明:算法具有好的不可感知性,能够抵御加噪声、重采样、重量化、低通滤波、MP3压缩等常规信号处理操作,以及抵御轻微的时域伸缩修改操作,尤其对于去同步的抖动攻击、随机剪切攻击具有很强的鲁棒性。

In view of copyright protection of audio works, the robust audio watermarking algorithm is proposed from the angle of the time domain. With the neighborhood averaging method, two proper positive integers are chosen to compute the mean of continuous sample values by sliding one sample every time. The two mean se- quences construct two mean lines which cross each other in the time domain. Then the watermark information is embedded at these cross points with the quantization index modulation （QIM） method. The experimental re- sults show the algorithm is imperceptible, and is resistant to common audio signal processing operations such as noise adding, re-sampling, re-quantization, low-pass filtering, MP3 compression etc. Moreover, it also sur- vives slight time-scale modification attack. Especially, it has strong robustness to random cropping and jitte- ring attacks.