数字语音拼接篡改检测技术研究
[Abstract]:With the development of the Internet, digital voice has gradually come into people's daily life. Not only can people communicate and communicate on the network through digital voice, but also voice information is developing rapidly as a medium of human-computer interaction. Digital speech, as a common multimedia information, brings us great convenience in our life. At the same time, with the continuous improvement of the functions of audio editing and processing software, it is easier and more convenient to edit and modify the voice. And the human ear is usually difficult to distinguish the traces of these treatments, if used illegally or spread maliciously, it will bring a serious threat to social stability. Therefore, for the authenticity of voice information, integrity verification becomes particularly important. In this paper, the key problems in the verification of the authenticity and integrity of digital speech information are studied and explored, and the related problems of digital speech dual compression detection and digital speech mosaic tampering detection technology are mainly studied. Specific research work can be divided into the following four aspects: 1. Digital speech tampering often leads to double compression, and the bit rate of double compression is the same as that of single compression. Therefore, this paper first studies the problem of double compression detection of digital speech at the same bit rate. Using the quantization spectrum coefficient in the process of MP3 coding as the feature, the variation between adjacent compression is analyzed, and the detection of double compression. 2. Double compression detection of digital speech also brings the problem of anti-forensics, that is, the dual compression speech is detected as one compression. In this paper, by studying the traces left by speech compression, a simple method of removing the traces of compression by adding zero sampling points is put forward, which realizes the research of anti-forensics with double compression. Through the study of this problem, it provides a strong support for the subsequent tamper detection. 3. The detection of digital speech splicing tampering is the focus of this research. Through the research of the double compression detection of speech, it is found that the tampering operation will also remove the compression trace. Therefore, the tamper localization is realized by using the inconsistency of compression times before and after tampering. The experimental results show that it has some effect, but the accuracy of the location is greatly affected by the accuracy of the location. The tamper localization method based on compression history inconsistency is not too high in the accuracy of location and positioning, in order to further improve the accuracy and accuracy of location. This paper also proposes to use the quantization characteristic of coding to locate the tamper. The zero-value spectrum coefficient is used as the feature to analyze the change features before and after quantization, and an effective feature is extracted to detect the tampering. The experimental results show that the detection accuracy is 99%.
【学位授予单位】:宁波大学
【学位级别】:硕士
【学位授予年份】:2017
【分类号】:TN912.3
【参考文献】
相关期刊论文 前10条
1 庄景晖;黄添强;;基于颜色特征的同源视频复制-粘贴篡改检测[J];闽南师范大学学报(自然科学版);2014年03期
2 袁秀娟;黄添强;苏立超;陈智文;吴铁浩;;基于边缘异常与压缩跟踪的视频抠像篡改检测[J];计算机工程;2014年07期
3 刘雨青;黄添强;;基于时空域能量可疑度的视频篡改检测与篡改区域定位[J];南京大学学报(自然科学);2014年01期
4 杨高波;龙永红;陈威兵;;基于最大似然估计的自适应阈值视频被动取证[J];湖南大学学报(自然科学版);2013年11期
5 陈智文;黄添强;吴铁浩;袁秀娟;苏伟峰;;同源视频Copy-Move篡改检测及恢复[J];计算机系统应用;2013年09期
6 刘育明;姚陈果;孙才新;袁智勇;Liu Yilu;;基于电网频率的数字录音真伪鉴别研究[J];仪器仪表学报;2013年06期
7 余先敏;王让定;严迪群;朱杰;;基于相同压缩速率下的MP3双压缩检测方法[J];计算机工程与应用;2013年12期
8 张静;宋怡;苏育挺;;基于空时域联合匹配的视频篡改检测算法[J];电子测量技术;2011年11期
9 黄添强;陈智文;苏立超;郑之;袁秀娟;;利用内容连续性的数字视频篡改检测[J];南京大学学报(自然科学版);2011年05期
10 孙君顶;马媛媛;;纹理特征研究综述[J];计算机系统应用;2010年06期
相关博士学位论文 前2条
1 张荣;数字图像真实性被动取证技术研究[D];宁波大学;2014年
2 严迪群;压缩域音频隐写与隐写分析中若干问题的研究[D];宁波大学;2012年
相关硕士学位论文 前4条
1 马朋飞;音频二次压缩分析方法与检测技术研究[D];宁波大学;2014年
2 余先敏;压缩域音频隐写分析技术研究[D];宁波大学;2013年
3 高锦;基于SVM的图像分类[D];西北大学;2010年
4 张力光;基于压缩域音频的信息隐藏技术研究[D];宁波大学;2009年
,本文编号:2192439
本文链接:https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2192439.html