Speaker Authentication Method using Reservoir Computing for Security System

Abstract

In recent years, we have been using biometric authentication systems in various places such as daily life and businesses. However, it's insufficient in hospital and food factory to introduction of the security system of the room access control. This is because they wear gloves, masks and hats in hospitals and factories, so they cannot authenticate faces or fingerprints. To solve this problem, I turned my attention to voice authentication. In this study, I propose a speaker authentication system based on Reservoir Computing. Reservoir computing is a new type of recursive neural network. In this study, we conducted classification experiments on 3, 5, and 10 speakers. The results show that the F-measure is above 0.9 for all the number of speakers.