Mobile device source identification method and system based on convolutional neural network

Abstract

The invention, which belongs to the technical field of speech evidence obtaining, discloses a mobile device source identification method and system based on a convolutional neural network. MFCC features for training a speech segment are extracted to train a GMM-UBM model; MFCC features are extracted based on a specific noisy speech segment and thus a GMM parameter is adjusted; and the extracted features are used for training a convolutional neural network and automatic recognition and classification are carried out. According to the invention, when the GMM-UBM model is trained, data are screened based on phonemes and tones of speech data and a few of representative data are selected, so that the representation generalization of the model is ensured, the data computation amount is reduced,and the modeling efficiency is improved. Since the GMM-UBM model is trained and then the GMM parameter is adjusted based on an MAP adaptive algorithm, a problem that the GMM model can not be trained because of a few of samples is solved and the computing is accelerated.

Description

technical field [0001] The invention belongs to the technical field of voice forensics, and in particular relates to a convolutional neural network-based mobile device source identification method and system. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: [0003] Device source identification is a detection method based on channel estimation of recording devices. In recent years, with the development of information technology, the source of digital audio data has become very convenient, and digital audio forensics technology has increasingly received widespread attention, especially in the judicial field. important application requirements. The detection of the device channel is mainly based on the noise characteristics of the device source. The noise can be divided into additive noise and product noise, that is, convolution noise. During the speech generation process, the channel will be accompanied by conv...

AI Technical Summary

Problems solved by technology

In the process of finding such a hyperplane, it is necessary to search for a hyperplane in a higher dimension, which brings a lot of inconvenience to the operation, reduces the speed of the operation, and also reduces the accuracy of recognition.

[0008] (2) At present, the number of devices in the data set for mobile device source identification research is relatively small, and the number of device sources used in this scheme is 21, which increases the technical difficulty of this scheme

In terms of justice, voice data is becoming more and more important as evidence, but some lawbreakers forge and tamper with voice data to conceal the truth in order to escape legal punishment, which brings a lot of trouble to investigators

Images