Study On The Two Kinds Of Digital Audio Forgery Detection

Recently the use of different multimedia devices have become very popular.Different kinds of portable recording devices are easily available because of the growing market trend and demand of different kinds of electronic gadgets.This fact has made it very easy for the transfer of audio materials from one place to another,from one person to another.In this world,where the availability of robust software is abundant for numerous purposes,there are many powerful digital audio editing software that can easily edit the audio files and bring about desired changes.These manipulated audio files when used in courts for judicial proceedings as evidence,for any sort of evidences in any cause and/or incidents may lead to undesirable consequences which may not be socially viable or may not be viable in any other form.So,it becomes very important to identify the integrity of the digital audio files and this fact supports and stands for the importance of audio forensics.Digital audio forgery detection techniques are being widely employed in different agencies to check the credibility,authenticity of the audio data.This research work basically focuses on the audio copy-move forgery detection and splicing forgery detection which are content-based forgery.The main work of this research work is as follows:1.The basic principle of digital audio forgery is discussed and explained;and general working and detection process for digital audio forgery detection techniques are explored.Two different methods for audio forgery detection were simulated and experimental results were analysed to identify the limitations.2.In this paper,for digital audio copy-move forgery in the same audio,a method based on distance measures to identify and locate the forgery is proposed.We focused on identifying the degree of similarity for audio features by using distance measures.Euclidean Distance and Chebycheb Distance are the distance measures used in the proposed method to identify the degree of similarity.The fundamental frequency,i.e.pitch is the audio features considered.Yet Another Algorithm for Pitch Tracking(YAAPT)algorithm is used for pitch detection.By calculating the distance and comparing with threshold value,we identified the copy-move forgeries in the audio file.The method can efficiently detect the copy-move forgery while speeding up the processing time and improving the accuracy while detecting the forgery.3.In this paper,a method based on channel response multi-feature for audio splicing detection and localization is proposed.Gaussian Mixture Model(GMM)on RASTA-MFCC coefficients is used for blind channel estimation.Channel response of the audio signal and dynamic and static information of the logarithmic spectral characteristics for the query signal is used to construct the feature to identify the tampering and locate.Sequence forward selection(SFS)algorithm for the purpose of filtering out the features is employed.At the time of location detection,we first determined whether the audio was tampered or not and then we take the average value of a number of starting audio frames that act as the reference features.The correlation coefficients of reference features and each of other audio frame features were calculated separately.On comparison with the existing algorithm,the results showed the algorithm improved the detection rate of forgery and their localization.