Digital audio tampered point detection method and system based on GMM-BIC

Abstract

The invention belongs to the technical field of digital audio signal processing, and discloses a digital audio tampered point detection method and system based on GMM-BIC. The method and system use aGMM-BIC method for digital audio tampered point detection instead of a traditional SGM-BIC method. In virtue of the characteristics of the GMM, the method and system can represent more information, and have a good representation effect on both long and short voice segments. A neural network is configured to classify fitted features to obtain a better classification effect. The method and system can automatically position the digital audio tampered points, have good adaptability, have good robustness to tampering operations of splicing close tampered points and concealing the tampered points byusing various noises, and provide a universal algorithm for the accuracy and the automation of digital audio tampered point detection.

Description

technical field [0001] The invention belongs to the technical field of digital audio signal processing, in particular to a GMM-BIC-based digital audio tamper point detection method and system. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: [0003] With the development of computer and Internet-related technologies, the way people exchange information is rapidly shifting from analog to digital. The advantages of easy storage, editing and dissemination of digital audio data bring a lot of convenience and fun to people's daily life. Most people can use audio editing software to splice, add noise and transform digital audio files simply and quickly without any professional knowledge. and so on. However, while the digital age brings convenience to people, it also brings hidden dangers, that is, the issue of digital information security. Digital audio tampering detection is an important branch of digital audio ...

AI Technical Summary

Problems solved by technology

[0006] 1) There is no unified digital audio tampering positioning standard operating procedure, and there is no unified evaluation standard for the experimental results;

[0007] 2) Most of the experimental methods have a large amount of calculation and the positioning is not accurate;

[0008] 3) The experimental scheme cannot be automated, and researchers need to rely on empirical knowledge to set the threshold as a criterion, and the adaptability to different database signals is poor

[0009] 4) At present, relevant research in the field is basically based on the premise that the tampering point is located in a silent frame (noise frame). On the one hand, research on tampering operations such as stitching, deletion, and copy-paste is more challenging and practical. More research work is focused on This type of tampering operation

For this kind of tampering operation, the purpose is to distort the semantics of the speaker, so the tampering point is located in the silent frame, which is less detectable and easier to operate; on the other hand, when the paper uses the silent frame as background noise, because the background noise Separation from speech is very difficult. If the detected tampering point is located in the speech frame, the possibility of false detection of this result is very high, and it is often not adopted.

Images