摘要
提出数字语音信号的FDCR(frequency domain coefficient residuals)特征,分析该特征对信号处理操作的鲁棒性和恶意攻击的脆弱性。将语音信号分帧、帧号映射为二进制的水印序列,通过量化FDCR特征的方法将水印嵌入到载体信号中。提取端通过判断提取帧号是否完整来对各帧内容进行取证,篡改定位。实验结果表明了该算法的不可听性、鲁棒性以及对于恶意攻击的篡改定位能力。
The FDCR(frequency domain coefficient residuals)feature of speech signals was proposed,and the robustness and fragility of this feature was analyzed.The speech signal was divided into frames,and the frame number was mapped to a binary watermark sequence,which was embedded in the carrier signal by quantizing the FDCR feature.For watermarked signal,the extractor performed content authentication by judging whether the extracted frame number was complete,and the attacked frame was located.Experimental results show that the proposed algorithm has good inaudibility and can tolerate a certain degree of signal processing operations.The attacked content can be located more accurately.
作者
师春灵
钱清
SHI Chun-ling;QIAN Qing(School of Big Data and Artificial Intelligence,Xinyang University,Xinyang 464000,China;School of Information,Guizhou University of Finance and Economics,Guiyang 550025,China)
出处
《计算机工程与设计》
北大核心
2021年第9期2455-2461,共7页
Computer Engineering and Design
基金
国家自然科学基金项目(61902085)。
关键词
数字水印
数字取证
数字语音
信号处理
篡改定位
digital watermark
digital forensics
digital speech
signal processing
tamper location
