Research On Splicing Detection For Digital Speech

Digital speech has gradually entered the people's daily life with the development of the Internet.People can through digital speech communication on the network,and speech information as a human-computer interaction media is rapidly developing.Digital speech as a common multimedia information bring great convenience to our lives,at the same time,audio editing software functions continue to improve.The speech editing and modification becomes easier and more convenient.Also,it is difficult for the human ear to distinguish the traces,if illegal use or malicious spread will bring serious threat for social stability.Therefore,the authenticity and integrity verification of the speech information becomes particularly important.In this paper,the key issues in the authenticity and integrity verification of digital speech information are studied and explored.The related problems of digital speech double compression detection and digital speech tampering detection are mainly studied.Specific research work can be divided into the following four aspects:Firstly,digital speech tampering often results in double compression process,and the double compression bit-rate is the same as single compression.Therefore,this paper firstly studies the double compression detection problem of the same bit-rate of digital speech.And we use the quantized spectral coefficients in the MP3 coding process as the characteristics,and analyze the variation characteristics of the two adjacent compressions and realize the detection of the double compression.Secondly,double compression detection of digital speech can also lead to anti-forensics problem,how to make double compression speech is detected as single compression.In this paper,we study the traces of speech compression,and propose a simple method by adding the zero value sampling to remove the traces of compression,so as to achieve the double compression of the anti-forensics research.Through the study of the problem for the tampering detection provides a strong support.Thirdly,the tampering detection of digital speech is the focus of this research.Through the study of the double compression detection of speech,it is found that the tampering operation will remove the compression traces.Therefore,this paper uses the inconsistency of the number of compression for tampering detection.The experimental results show that the method can effectively detect tampering speech,but the detection accuracy is greatly affected by the positioning accuracy.Fourthly,the tamper localization method based on compression history inconsistency is not too high in detection accuracy and positioning accuracy.This paper also proposed usingquantization artifacts for tampering localization detection.The zero values of spectrum coefficients as the feature analysis of the changing features for quantization process.An effective feature was extracted to detect tampering,the experimental results show that the detection accuracy reached 99%.