摘要
提出了一种复倒谱域的自适应音频盲数字水印技术,水印嵌入过程在音频的复倒谱域中进行,并采用循环嵌入,嵌入的强度由实际应用中检测过程的具体要求控制。该文采用原始音频的一个或两个特征参数估计引起失步的参数,在水印检测前利用估计的参数对其进行同步恢复,重新获得检测和嵌入的同步。水印采用的是具有实际意义的音频。检测过程中采用独立分量分析技术,在不需要任何原始音频、水印、嵌入的信息以及攻击等信息的情况下,可以检测到水印并可精确提取水印,实现了真正意义上的盲检测。实验数据说明这种水印算法具有很好的鲁棒性。
This paper proposes an adaptive blind audio watermark technique based on complex cesptrum domain. Watermark is embedded iteratively, and the embedded intensity is controlled by the requirements of the application. A new method to estimate the attacks parameters by using the characteristics of the original audio is proposed. By the estimated parameters the detector can re-obtain the synchronization between embedding process and detection process. The watermark is meaningful audio. The detector adopts the ICA, which can not merely extract but detect the watermark without any information about original audio, watermark, embedding process and attacks. The experimental results show that the proposed watermark technique has a good robustness against audio manipulations.
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第4期128-130,133,共4页
Computer Engineering
基金
国家自然科学基金资助项目(60502027)
深圳市科技计划基金资助项目(200336)
关键词
自同步
独立分量分析
复倒谱
音频盲水印技术
Self-synchronization
Independent component analysis（ICA）
Complex cesptrum
Blind audio watermark
