Forgery Detection Of Digital Audio

With the development of the computer and network technology, digital multimedia had been widely used. Compared with traditional way, digital multimedia technology has many advantages: easy to store, easy to edit and easy to transmit, which also caused a lot of new problems.Firstly, one can perform forgeries to digital file by low-cost ways. With the help of some kind of software, an image, a piece of video or audio can be tampered easily. Secondly, we cannot judge whether a digital file has been tampered, which will bring us some trouble while using the digital multimedia technology, especially in some special scene. For example, judicial evidence, confidential documents, medical records, historical documents backup, etc. It is undoubtedly very significant to ensure the authenticity and integrity of such information.At present, a number of technologies such as digital watermarking and digital signatures could protect the authenticity of the digital data very well. However, this requires the device itself insert watermark and signature when taking of the original data. Such equipment is usually expensive, and their utilization rate is very low, most of the equipment do not have such functionality.Therefore, the research topic is how to perform the authenticity test for a given digital audio material which was not provided additional significant information (i.e., without any supporting information to ensure its integrity and authenticity).With the use of wavelet noise reduction method, we can get the digital audio noise. Then this paper studied the noise characteristics, according to its characteristics of energy distribution, we use a second filter to extract its features, which is regarded as the characteristics of the background noise. And finally this paper finds a way to detect whether a piece of audio has been tampered by another piece of audio recorded by another type of device. Secondly, this paper studied the characteristics of voice signals, tried to extract the vocal fragments by the way of combining short-term energy analysis and short-term zero-crossing analysis, and put forward a way to detect the authenticity of the digital audio by comparing the similarity of these fragments.