Playback speech detection method

Abstract

The invention discloses a playback speech detection method. At a training stage, a first variation coefficient vector, a normalized first cepstrum feature matrix, a second variation coefficient vectorand a normalized second cepstrum feature matrix of each speech sample in a speech database are firstly acquired as four kinds of features; the four kinds of features of all positive samples are theninputted to a GMM model for training, four positive sample feature models are obtained, and besides, four negative sample feature models are obtained; at a testing stage, four kinds of features for to-be-detected speech are acquired in the same way, the kinds of features are inputted to the corresponding positive sample feature models and the negative sample feature models respectively, and four likelihood ratio scores are obtained; and according to the four likelihood ratio scores, a final score is obtained, and through compairing the final score and a judgment threshold, whether to be a playback speech is judged. The playback speech detection method is not only limited to a text-related voiceprint authentication system, and has the advantages of low equal error detection possibility, strong robustness and relatively low calculation complexity.

Description

technical field [0001] The invention relates to a voice detection technology, in particular to a playback voice detection method. Background technique [0002] In the field of biometric technology, the voiceprint recognition system has been widely used in the fields of life, finance and justice due to its high security and convenient acquisition. With the continuous development of voiceprint recognition technology, the attacks of various counterfeit voices on the voiceprint recognition system are becoming more and more severe. In the past few years, researchers have focused on the detection of counterfeit voice mainly on synthetic voice and converted voice, and to some extent ignored the attack of playback voice on voiceprint recognition system. In fact, first of all, since the playback voice is directly recorded from the real voice, it is more threatening than the synthetic voice and converted voice; second, the playback voice is more convenient to obtain than other counte...

AI Technical Summary

Problems solved by technology

In the early research work, Zhang Lipeng and others proposed a detection algorithm based on the silent segment of speech by modeling the silent segment of speech. Through the research on the algorithm, it was found that although the algorithm has certain performance in the detection of playback speech, the algorithm Because of the small range of silent segments used in this method, it is easy to be polluted by background noise, so the algorithm has certain limitations.

However, Wang Zhifeng and others proposed a recording playback detection algorithm based on channel mode noise by exploring the mechanism of playback voice generation and analyzing the noise differences introduced by different devices. Although this algorithm has a good detection effect, its experiments only use A recording device and a playback device, so the robustness of the algorithm needs to be studied

In addition, foreign research such as Shang and Stevenson proposed an algorithm to detect the similarity between the test voice and the legal voice on the peak map by using the randomness of voice generation. This algorithm can only be applied to text-related voiceprint authentication. system

On the basis of an algorithm for detecting the similarity between the speech to be tested and the legal speech on the peak map, Jakub Galka et al. added the positional relationship of each frequency point to the characteristics of the peak map, which further improved the anti-replay of the voiceprint authentication system. Voice performance, but the algorithm is still limited to text-related voiceprint authentication systems

In recent years, Todisco M and Delgado H et al. proposed a CQCC (Constant Q Cepstral Coefficients) feature based on constant Q transform in 2016. Although it has a certain effect on the detection of playback speech, the detection accuracy still needs to be improved.

Ji Z and others used a variety of combined features and integrated classifier ideas to reduce the equal error probability of playback speech detection to about 20%, but the complexity of the algorithm is extremely high

Lantian Li et al. proposed the I-MFCC method for playback speech detection by using the F-ratio method to analyze the difference between real speech and playback speech. Experiments show that although this feature has a certain detection effect, the robustness of this feature Poor sex

Images