Study On The Authenticity Identification Of Digital Audio

As the rapid development of digital audio, the application prospect of digital audio is becoming widespread. When applied to some special occasions, the truth of digital audio is significant. For example, as the electronic evidence in court, the authentication of digital audio is crucial for the court decision. This need makes it urgent to develop and perfect the authenticity identification of digital audio.This paper mainly studies the digital audio authenticity identification technology. It proposes three detection methods to the possible forgeries. The main structure of the thesis is as follows:(1)According to the re-quantization of digital audio, the detection scheme based on Benford's Law is proposed. This algorithm counts the probability distribution of the first significant digit of sub-frame DCT coefficients. Then the probability distribution is compared with the one predicted by Benford's law. For the original audio, the fitting is very good while that of re-quantified audio is not. So the re-quantified audio can be detected by this principle.(2) According to the duplication forgery of digital audio, this paper presents a detection method that can not only detect the simple duplication forgery but also the duplication forgery with palliative operation. This method can detect the simple duplication forgery and the one with amplitude scaling and translation through difference and multiple sequences, while detecting duplication forgery adding noise by Principal Component Analysis.(3)According to the up-sample forgery of digital audio, this paper proposes a detection method based on spectral distribution. Up-sample increases the sampling rate statement of digital audio, but can't increase the frequency information, while the high frequency distribution is less and more smooth. So this method can detect the up-sample and estimate the original sampling rate through the analysis of the distribution of the audio frequency.Experimental results show that these three methods can meet the demand of detection of digital audio. The re-quantization, duplication and up-sample can be detected effectively.