Design And Implementation Of Voiceprint Recognition System Based On Embedded

The growth of science and technology has made people’s lives simpler,but it has also forced them to pay attention to the protection of personal identifying information.Because of its great stability,high security,and minimal privacy infringement,voiceprint recognition is a frequently utilized identity authentication security technique in the field of identity information security.Nowadays,the research of voiceprint recognition technology is oriented to practical development.Due to the convenience and operability of embedded devices,apply voiceprint recognition to the embedded field,the development of voiceprint recognition technology applications for embedded devices has practical implications.The main work of the paper:(1)The Mel frequency cepstrum coefficient and Gammatone frequency cepstrum coefficient are fused to enhance the approach in the feature extraction stage of voiceprint recognition,in order to tackle the problem that single feature extraction parameters cannot properly capture speech features.And for the problem of increasing the amount of calculation and data redundancy caused by the fusion of multiple feature parameters,this paper uses Fisher criterion to reduce the dimension,and finally forms a new feature extraction parameter MFCC-GFCC.(2)In the modeling phase of voiceprint recognition,in order to optimize the selection of the initial point of the traditional Gaussian mixture model,the Euclidean distance-based K-Means clustering algorithm is combined with the traditional Gaussian mixture model to form a new improved model as a modeling method for voiceprint recognition.(3)The enhanced technique is empirically confirmed in the two steps of feature extraction and model establishment,and the overall performance of the improved algorithm is examined.The experiment reveals that the modified method can improve the system’s recognition rate,increase its resilience,and reduce its recognition time.(4)Design and run the system on Raspberry Pi 4B development board.Firstly,the voiceprint recognition system is designed based on the improved algorithm after verification,and the program design of the four main stages of pre-processing,feature extraction,model establishment and matching discrimination is introduced.Then the user information module,selection module,training module and recognition module of the system are described in depth.Finally,the entire system is put through its paces to ensure that it can perform voiceprint recognition and provide correct feedback on the findings.This paper’s algorithm verification and system design are done on embedded devices.The accuracy and ease of the voiceprint recognition system have substantially improved thanks to the enhanced voiceprint recognition algorithm,establishing the groundwork for the use of voiceprint recognition technology in the embedded industry.This paper’s research has some implications for the development of voiceprint recognition systems based on embedded systems.