Evaluation of biometric user authentication using an ensemble classifier with face and voice recognition

TL;DR

This paper develops and evaluates a bimodal biometric authentication system combining face and voice recognition, demonstrating high accuracy and low error rates across multiple benchmark datasets.

Contribution

It introduces an ensemble framework integrating face and voice classifiers and evaluates its performance on standard datasets, showing significant accuracy improvements.

Findings

Achieved over 99% accuracy, precision, and true positive rates.

False positive and negative rates below 1%.

Demonstrated robustness across multiple benchmark datasets.

Abstract

This paper presents a biometric user authentication system based on an ensemble design that employs face and voice recognition classifiers. The design approach entails development and performance evaluation of individual classifiers for face and voice recognition and subsequent integration of the two within an ensemble framework. Performance evaluation employed three benchmark datasets, which are NIST Feret face, Yale Extended face, and ELSDSR voice. Performance evaluation of the ensemble design on the three benchmark datasets indicates that the bimodal authentication system offers significant improvements for accuracy, precision, true negative rate, and true positive rate metrics at or above 99% while generating minimal false positive and negative rates of less than 1%.