Human voice identification based on the detection of fundamental harmonics

The article is devoted to the development of a system for biometric identification of a person by voice. The article discusses algorithms for the analysis of audio recordings for biometric identification of a person by voice. The technique of experimental research is considered, the process of processing the results of identification is described. Used algorithms MFCC and PLP for digital processing and analysis of audio recordings. An algorithm based on multi-criteria optimization has been developed for acoustic speech analysis. Various algorithms are used, such as hidden Markov models (CMM or HMM), as well as a model of a mixture of Gaussian distributions (GMM or GMM); in recent years, Wave Net neural networks have been actively used. The result of determining the tone of speech and the content of speech for the purposes of identification by voice is obtained.In the work, to compare the recorded voice with the saved voice for the purpose of personal identification, an unlimited text independent recognition system was applied using the Gaussian mixture model. The recorded voices were processed and stored during the registration phase, and the probing voices were used for comparison during the verification/ recognition phase of the system. For biometric identification of a person by voice, the MFCC and PLP algorithms were used for digital processing and analysis of audio recordings. The result obtained makes it possible to determine the fundamental harmonics of speech for the purposes of identification by voice. The “Multiparameter automated system of biometric identification of a person” was developed on the Visual FoxPro DBMS. Today, speech identification and authentication is used in a wide range of applications ranging from smartphone applications to access control systems. Additional confirmation of the relevance of this area is the many research centers.