Access Control Using Specific Code and Biometric Identification

Abstract

This study aims to develop a verification and identification system for speakers with intelligent speaker recognition, by relying on MFCC and PLP algorithms coupled with ML models like SVM, Random Forest, and Neural Networks. The system was then tested on a database of 24 speakers, where SVM, followed by Neural Networks and Random Forests, showed best results with PLP features, while Gradient Boosting showed poor results. The study recommends increasing the database, implementing data augmentation techniques, testing the models in real-life scenarios, and combining voice with other biometrics for enhanced security.