Speaker recognition method based on Triplet-Loss

Abstract

The invention relates to a speaker recognition method based on Triplet-Loss. The speaker recognition method comprises the following steps: S1, acquiring voice signals, comprising three groups of samples, respectively a group of voice sequence of a speaker, another group of voice sequence of the same speaker and a group of voice sequence of a different speaker; S2, carrying out preprocessing on thevoice signals, so that channel noises generated during the voice collecting process are removed; S3, carrying out voice feature parameter extraction on the voice signals after denoising; S4, constructing an RNN neural network based on an LSTM neural network; S5, training the RNN neural network by adopting 90% of the extracted three groups of voice feature parameters as the input of the RNN neuralnetwork; and S6, after the RNN neural network is trained, carrying out speaker recognition by taking the rest 10% of the three groups of voice feature parameters as the input of the RNN neural network. The speaker recognition method has the advantages that the accuracy is high, the recognition effect is good, the reliability high, etc.

Description

technical field [0001] The present invention relates to the technical field of neural network and deep learning, in particular to a speaker recognition method based on Triplet-Loss. Background technique [0002] As information security issues become more and more serious, the impact is growing. The issue of "personal privacy and confidentiality" urgently needs to be resolved. How to accurately and safely determine a person's identity arouses people's thinking. As a key interface of human-computer interaction, voice plays an important role in identity authentication. Voiceprint recognition is speaker recognition. As a unique biological feature of a speaker, voiceprint is a new means to overcome traditional authentication methods. Compared with other methods, the acquisition of speech containing voiceprint features is convenient and natural, and the voiceprint extraction can be completed unconsciously, so the user's acceptance is also high; the cost of acquiring speech reco...

AI Technical Summary

Problems solved by technology

[0003] Common voiceprint recognition methods, such as the early voiceprint recognition method based on signal processing, use some technical methods in signal processing technology to calculate the parameters of voice data in signal science, and then perform template matching and statistical analysis of variance. The data is extremely sensitive, the accuracy rate is very low, and the recognition effect is very unsatisfactory

[0004] The recognition method based on the Gaussian mixture model can achieve good results and is simple and flexible, but it requires a large amount of voice data and is very sensitive to channel environmental noise, which cannot meet the requirements of real scenarios.

[0005] The existing deep learning neural network-based methods do not consider the context-dependent nature of the speech signal, and the extracted features cannot represent the speaker well, and do not fully utilize the advantages of deep learning

Images