Automatic detection method of digital audio tampering based on power grid frequency fluctuation supervector

Abstract

The invention belongs to the technical field of digital audio signal processing, and discloses a digital audio tampering automatic detection method based on grid frequency fluctuation supervector, which analyzes the phase spectrum and instantaneous frequency spectrum sensitive to signal truncation in the grid frequency (ENF) signal, and extracts them respectively Effective feature set, and fuse the extracted feature set; only use a large number of original speech signals, including speech signals with various signal-to-noise ratios, and even some defective speech signals for background modeling, which is different from actual detection The situation is more consistent, so the background model is not sensitive to the type of signal tampering, and can effectively detect various types of tampered audio. The invention establishes a consistency model of the same type of speech signal, and filters out a large number of features irrelevant to this type of attribute through self-adaptation, and the self-adaptive part can also be adjusted by the user, which has better robustness.

Description

technical field [0001] The invention belongs to the technical field of digital audio signal processing, and in particular relates to an automatic detection method for digital audio tampering based on a grid frequency fluctuation supervector. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: [0003] In recent years, the rapid development of digital media technology and Internet information technology has made the application of digital media signals more and more extensive and frequent. At the same time, digital audio signal has also become one of the most popular multimedia applications, and its advantages of being easy to save, edit and spread bring a lot of convenience and fun to people's daily life. With the development of digital audio signal processing technology and the abundance of various easy-to-operate voice editing tools, it becomes easy to tamper and forge digital voice content. The prosperity of...

AI Technical Summary

Problems solved by technology

The prosperity of network-wide media products brought about by the integration of the three networks has accelerated the spread of digital audio and aggravated the hidden dangers of multimedia information security.

[0009] (1) Most of the methods are poor in robustness, and there are requirements for the quality of the signal to be tested and the recording environment, etc., and there is no consistent evaluation standard for the test results;

[0010] (2) Some detection methods require the experience or domain knowledge of professionals to judge whether the voice signal has been tampered with, and cannot be automated;

[0011] (3) Most of the current heuristic research schemes have poor adaptability, and there are various problems in practical applications, which cannot meet the requirements of digital audio blind tampering detection

[0013] At present, the definition of blind tampering detection of digital audio is not perfect enough, and there is no unified evaluation standard for relevant research results, so that follow-up research has not been followed up in time. This invention uses machine learning methods to weaken the operation process of tampering detection and emphasizes the use of the last The classification result is used as the efficiency of the invention, and it is automated; in order to ensure that the invention has a wider range of applications, the empirical behavior of threshold selection must be eliminated. The present invention uses the advantages of machine learning methods to process large data. The process is handed over to SVM training; the fundamental goal of digital audio blind tampering detection is that the blind tampering scheme can be applied to various databases and various scenarios. In order to ensure the application, the detection scheme must be robust in various practical situations

Images