Comparison of Authentication Systems Based on a Cluster of Algorithms with Polyphonic and Monophonic Types of Biometric Images

The article is devoted to the choice of architecture of the authentication module according to formant characteristics. For comparison, two systems of biometric authentication by voice are proposed: the first uses a biometric image that includes the formant characteristics of three vowel sounds (polysonic type of biometric image); the second interacts with three biometric images, each of which consists of the formant characteristics of a certain vowel sound (a monosonic type of biometric image). The operation of each biometric authentication system discussed in the article is based on a cluster. The cluster includes the Random Forest algorithm, k nearest neighbors, logistic regression, multilayer perceptron, and support vector machines. The architectures of authentication systems are given. The authentication process in each biometric system is presented step by step. A scheme for the interaction of algorithms in learning and authentication modes is proposed. The scheme for creating training samples is shown. The stages of creating training samples are given. An experiment was conducted with seven people. The results are analyzed.