An automatic detection method for digital audio tampering based on feature fusion

Abstract

The invention provides a digital audio tamper automatic detection method based on feature fusion. The method comprises the steps that a signal to be measured is downsampled; the standard frequency of electric network frequency (ENF) is used as the center to carry out band-pass filtering to acquire an ENF component in a signal; the phase feature based on DFT0, the phase feature based on DFT1 and the instantaneous frequency feature based on Hilbert transform are respectively extracted from the ENF component; feature fusion is carried out to acquire a feature set; finally, an optimized support vector machine classifier is used to train a part of data in the feature set to acquire a training model; and the training model can be used to predict a speech signal to be tested. According to the invention, representative phases and instantaneous frequency features in an ENF signal are used to carry out feature fusion; the support vector machine is used for classifying to acquire a classification model; and the model can well detect signal insertion and deleting, and is more visual and simpler than a traditional discriminated method.

Description

technical field [0001] The invention relates to the technical field of digital audio signal processing, in particular to an automatic detection method for digital audio tampering based on feature fusion. Background technique [0002] Compared with analog audio signals, digital audio signals have advantages such as easy transmission, copying, and storage, but these advantages bring various conveniences to people, but also bring problems that cannot be ignored. That is, tampering with digital audio signals becomes just as easy and quick. The authenticity and integrity of digital audio are threatened from various aspects, and most tampering operations are difficult to judge by human senses. Digital audio forensics technology is a technology used to verify the authenticity and integrity of digital audio signals. Passive digital audio forensics is an important branch of digital audio forensics, and it is widely used in judicial forensics, news justice and scientific discovery. ...

AI Technical Summary

Problems solved by technology

[0005] It can be seen that there are some problems in the current research on passive tampering detection of digital audio based on ENF: 1) There is no authoritative ENF comparison database

Use the ENF component in the signal to be tested to compare with the ENF database to judge whether the speech signal has been tampered with, and there is no reliable result. 2) Part of the detection method is to judge whether the speech signal has been tampered with intuitively and visually, and cannot be automated.

3) At present, most of the detection methods are relatively simple, do not include more actual detection situations, and cannot meet the requirements of digital audio passive detection

Images