一种能进行篡改定位的音频双水印算法

摘要

提出一种音频双水印算法,解决双水印中认证水印的抗攻击鲁棒性差以及脆弱水印定位不准确的问题.鲁棒水印采用基于RBF神经网络的音频零水印方案作为版权认证水印,因为该水印方案不改变原始音频数据,所以具有良好的透明性;脆弱水印采用一种新的双极性量化方法嵌入,提高了脆弱水印的敏感性.实验结果表明,该方案认证水印具有很强的抗攻击鲁棒性,脆弱水印具有很强的敏感性,且能够对恶意篡改进行精确定位.%Aiming at the problems of dual watermarking algorithm as poor robustness and inaccurate positioning, a digital audio dual watermarking was presented.A zero-watermarking scheme system based on RBF neural network was constructed as robust watermarking.Since the scheme did not change the original audio data, it had a good transparency.The fragile watermarking embedded watermark image block by double bipolar, it had a good sensitivity.The experiment shows that the watermarks are robust and sensitive to many signal operations.Moreover, it has a good accuracy of tamper location.