CAJ | 학술논문

提出一种用于内容认证的脆弱音频水印算法.与以往的脆弱音频水印算法相比,不采用二值图像作为脆弱水印,而是基于音频内容,利用奇异值分解生成脆弱水印.水印嵌入在离散小波变换和离散余弦变换构成的混合域.仿真实验表明,该算法具有很好的不可感知性,并且对有损压缩、添加噪声、重量化等操作具有很强的敏感性,同时能够对替换等恶意篡改做出报警并确定篡改位置.
제출일충용우내용인증적취약음빈수인산법.여이왕적취약음빈수인산법상비,불채용이치도상작위취약수인,이시기우음빈내용,이용기이치분해생성취약수인.수인감입재리산소파변환화리산여현변환구성적혼합역.방진실험표명,해산법구유흔호적불가감지성,병차대유손압축、첨가조성、중양화등조작구유흔강적민감성,동시능구대체환등악의찬개주출보경병학정찬개위치.
A novel fragile audio watermarking scheme for content authentication was proposed in this paper. Compared with traditional fragile audio watermarking schemes, instead of using the binary image as the fragile watermark, singular value decomposition(SVD) was adopted to generate the fragile watermark based on the audio content in this scheme. The generated fragile watermark was embedded in the hybrid domain. Experimental results show that our fragile watermarking scheme is not only inaudible and sensitive against the MP3 attack, additive noise, re-quantization, etc, but also fragile to malicious modification. Especially, it can detect and localize the tampered regions effectively.