Method and system for detecting voice spoofing attack of speakers on basis of deep learning

Abstract

The invention discloses a method and a system for detecting voice spoofing attack of speakers on the basis of deep learning. The method includes constructing audio-frequency training sets, initializing deep feed-forward neural networks and deep recurrent neural networks and respectively training the deep feed-forward neural networks and the deep recurrent neural networks by the aid of multi-frame feature vectors and single-frame vector sequences of the training sets; respectively leading frame level and sequence level feature vectors of to-be-tested audio frequencies into two trained linear differential analysis models in test phases, weighting two obtained result grades to obtain scores and comparing the scores to predefined threshold values so as to discriminate voice spoofing. The method and the system have the advantages that local features can be captured, and global information can be grasped; the linear differential analysis models are used as classifiers in identification and verification phases, the voice spoofing attack can be judged by means of grade fusion, and accordingly the voice spoofing detection accuracy can be greatly improved.

Description

technical field [0001] The present invention relates to a technology in the field of intelligent speech, in particular to a method and system for detecting speaker speech spoofing attacks based on deep learning. Background technique [0002] Speech spoofing attack refers to the technology of forging the voice of a specific target to attack the automatic speaker recognition system. Speaker recognition technology has been widely used in many fields, such as: identity verification, Internet security, human-computer interaction, banking and securities systems, military criminal investigation, etc. In recent years, attacks on speaker recognition systems are mainly divided into four categories, namely imitation attacks, recording playback, speech synthesis, and speech conversion. Research shows that the main problem of traditional speech spoofing attack detection lies in feature extraction, and existing feature extraction methods have many shortcomings in terms of expressiveness ...

AI Technical Summary

Problems solved by technology

[0005] The present invention aims at the existing traditional voice spoofing attack detection method, which has the limitations of feature extraction that cannot accurately characterize the discriminative features between spoofed voice and real voice, dynamic features of lost voice signals, poor robustness, and poor recognition effect. In the feature extraction stage, the deep learning model is used to extract the feature vector, two different frameworks: frame-level feature representation based on deep feed-forward neural network And the sequence-level feature representation based on deep recurrent neural network, which can not only capture local features, but also grasp global information

Images