Design And Implementation Of Voiceprint Recognition System Based On ?-Vector

With the development of computer science and technology,in recent years,the development of network information security appears more and more frequently in daily life,more and more people are concerned about this issue.Because of the uniqueness and validity of the biometric recognition method,it has gradually emerged and attracted the attention of many academics and industry.Voice print recognition is a new technology for identity authentication based on speaker's voice,which enables users to fully participate in the process,and the cost of equipment required in the process is also low,making this technology can be widely used in finance,military,information electronics,Internet,public security and other fields.Meyer Frequency Cepstrum Parameter(MFCC)makes full use of the advantages of human physiological characteristics and the spectral characteristics of speech signals,and is often used as a feature parameter to describe the speaker's characteristics.In practice,one of the biggest problems encountered in the development of voiceprint recognition is how to overcome the inconsistency of voice data channels.To solve this problem,this paper adopts the i-vector/PLDA method as the core algorithm of voiceprint recognition,and compensates the I-vector features including speaker characteristics and channel information through PLDA to eliminate the interference of channel inconsistency.Because the pre-processing stage is very important for the final performance of the system,in the speech pre-processing stage,the method of combining short-term energy and LTSD voice endpoint detection is used to extract the effective speech segments.Before the PLDA stage,the system whitened the I-vector features with Gauss whitening,and the final effect has been improved to a certain extent.Experiments were carried out to compare the systematicness of different scoring methods.In order to improve the response speed of the system,the time-consuming part is optimized to reduce the time consumption in training and recognition,and the effect of system recognition is improved significantly.According to the development steps of standard software engineering,this paper makes a detailed analysis and design of the system,and realizes various functions of the system.The actual test shows that the voiceprint recognition system can not only meet the various needs of recognition,but also ensure the real-time and accuracy of detection in terms of performance.