Tamper Detection Of Digital Audios

With the development of computer technology and communication technology,multimedia data such as audios,videos and images have gradually become widely used.With those kind of message introduced into computer,the information carried can be greatly expanded,but also brings many new problems.Digital media processing technologies develop so fast,and the tampering technologies have much lower threshold,all making the audio tampering happens more often,and more difficult to detect.How to identify those digital multimedia files' originality,authenticity and integrality is becoming very challenging.This paper mainly deals with the identification of audio recording device and the detection of audio splicing.For the former one,this paper proposes that the audio signal is the sum of speech signal and noise signal,and the noise can be regarded as the intrinsic fingerprint traces of an audio recording device.These digital traces can be characterized and identified by new machine learning techniques.Thus a method based on background noise using deep learning technology is proposal,and the test result is pretty good.For the latter one,this paper further proposes a method of audio splicing detection through the model generated formerly.Also,several more deep learning methods are used to enlarge the audio splicing detection's range of application.To improve the accuracy,several models such as averaging model and voting model are also generated,and proved to be pretty effective.