Software Design And Implementation Of Voiceprint Recognition Module Based On ARM

With the popularity of mobile Internet applications,identity authentication has become a basic requirement for information security.Biometrics is gaining more and more attention because of its safety,speed and convenience.Voiceprint recognition is a kind of biometric recognition,and it is widely used in the fields of mobile finance,public security,justice and intelligent security because of its advantages of safety,convenience and efficiency.With the development of the Internet of Things,embedded electronic products have been widely used in various occasions due to their portability and economy,and they have become an important part of people's lives.Therefore,it is of practical significance to realize voiceprint recognition on embedded platforms.In this thesis,the Gaussian Mixture Model voiceprint recognition algorithm is improved and optimized to improve the performance of the algorithm,and the algorithm is implemented on the ARM embedded platform.The main innovation of the thesis lies in the in-depth study of the Gaussian Mixture Model algorithm,and an improved scheme of Gaussian Mixture Model initialization method based on distance probability is proposed for the local convergence problem of traditional Gaussian Mixture Model initialization.What's more,the algorithm is implemented on the ARM embedded platform.The main work includes: The related algorithms involved in voiceprint recognition are studied;In terms of feature extraction,Mel Frequency Cepstral Coefficients and Linear Prediction Coefficients are analyzed in detail,and then the two parameters are fused to form a new fusion feature parameter MFCC-LPC,which improves the accuracy of voiceprint recognition;In terms of recognition algorithm,for the local convergence problem of traditional Gaussian Mixture Model initialization,a Gaussian Mixture Model initialization based on distance probability is proposed,which uses the distance scale and probability idea to select the initial point,in this way,the accuracy of voiceprint recognition will be improved,and the training time of the model will be reduced;Finally,the improved voiceprint recognition algorithm is implemented on the ARM embedded platform and the human-computer interface of voiceprint recognition is designed.Experiments were carried out by setting up an experimental platform.The results show that the recognition accuracy of the improved voiceprint recognition algorithm is increased by 0.4% and the training time is decreased by 10%.For 25 seconds of training speech and 5 seconds of test speech,the recognition accuracy is 97.67%,the recognition time is 287 milliseconds,and the training time is 14.4 seconds,which satisfy the design specifications of this thesis.The ARM-based voiceprint recognition module designed in this thesis has the advantages of real-time,accuracy,portability and good user interface,which lay a foundation for the future work of voiceprint recognition.