Research On Digital Audio Copy-move And Splicing Detection

With the rapid development of digital multimedia technology and the gradual popularization of the Internet,it is becoming more and more convenient for people to share digital media information.At the same time,with the rise in various editing software it has become easy to manipulate digital multimedia files making its authenticity and integrity often in doubt.Especially digital audio files,as one of the widely used multimedia files,have important security implications for researching audio forensics technology.Based on the type of audio semantic tampering,this paper proposes three detection algorithms for homologous copying and tampering,and a detection algorithm for heterogeneity tampering.The main work of this paper is as follows:1.The typical detection algorithms of homologous audio copy-move tampering and heterogeneous audio splicing tampering are studied,and their simulation analysis is performed respectively.The deficiency of the algorithm is pointed out.2.For the homologous copy-move tampering,two tampering detection algorithms based on time domain features are proposed.The first method is based on the detection algorithm of content matching degree,selects the appropriate length to divide the audio syllables into separate segments,and compares the similarity of small audio segments between pair of syllables,if two pairs of segments are larger than the set threshold then it is judged as tampered.The second method is based on the fast detection algorithm of syllables using time domain histogram features,and the time complexity is reduced from the same algorithm from O()to O(log)n.Firstly,the feature of histogram characterizing of each syllable is extracted,and the correlation coefficient of the two-syllables histogram feature in the neighbouring mean is calculated after the mean value of the histogram is quickly sorted.When the correlation coefficient is greater than the set threshold,two syllables are determined to have replication relationship.Experiments show that both algorithms have better detection performance.3.For the homologous copy-move tampering,a tamper detection algorithm based on spectral feature inter-frame correlation is proposed.First zero-cross rate pre-judgment is used to measure audio,then calculate the amplitude spectrum similarity between pair of syllables and then match all the pairs to identify the pairs that have the copy-paste relationship according to the set conditions and determines the tampering area.Experiments show that the spectral features are better than those of other transform domains,and they are robust to noise addition,resampling,and compression.4.For the heterogeneous audio splicing tampering,a detection method based on MelFrequency Cepstrum Coefficients(MFCC)is proposed.Firstly,the query audio is denoised by Wavelet Transform(WT),and the MFCC coefficients are calculated separately from the query audio and noise-reduced signals.Then the background noise of the device is obtained,and the stability of the feature is discussed.Finally,the sliding window is used to calculate the correlation coefficient vector of the adjacent audio features,and the threshold value is determined according to the value of the correlation coefficient vector and its first-order difference.It is determined whether there is tampering in the query audio and the tampering position is determined.Experiments show that the algorithm is robust to resampling and compression.