Voice Recognition System for User Authentication Using Gaussian Mixture Model

The use of biometrics in the user authentication process is the leading choice today. One of the biometrics that can be used is the human voice. In this paper, a voice authentication system using the Gaussian Mixture Model (GMM) is proposed. GMM was chosen because of the ease and accuracy in classifying the data. Voice data features are extracted using Linear Predictive Coding (LPC) before being classified using GMM. Voice data was recorded directly from 30 respondents using laptops and smartphones. Additional devices in the form of earphones were added to get better results. The system's learning process has an accuracy of 84%, and the overall testing process has an accuracy of 82 %. There are also differences in the accuracy of user authentication between data that use enhancements and those that do not. They are 87% and 72 %, respectively.