摘要
根据水印调整直流电平偏移的音频水印算法具有很好的低通和抗噪性能。当水印发生去同步时,如果同步点偏移不大,直流电平的符号不会改变,但当同步点偏移很大时,直流电平可能变号,从而导致水印提取错误。为此,提出了在宿主音频中嵌入同步训练序列的抵制去同步攻击的改进算法。实验结果表明,算法对随机裁剪等去同步攻击的鲁棒性得到了明显提高。
Audio watermarking that modulates DC level shift according to watermark has wonderful low pass and noise resistance performance. When loss of watermarking synchronization occurs, if the shift of synchronization position is trivial, the sign of DC level will not be changed, but if notable, the sign of DC level will be changed, as result watermarking extracting error occurs. In order to resist de-synchronization attacks, an improved algorithm that embeds training sequence in cover audio is proposed. Experimental results show that this technique obviously improves the watermarking robustness resistance to de-synchronization attacks.
出处
《工程图学学报》
CSCD
北大核心
2006年第3期161-166,共6页
Journal of Engineering Graphics
基金
国家自然科学基金资助项目(60572096)
浙江省自然科学基金资助项目(Y104144)
宁波市博士基金资助项目(2005A610003)
关键词
计算机应用
音频水印
直流电平
训练序列
computer application
audio watermarking
DC level
training sequence