Theoretic evidence k-nearest neighbourhood classifiers in a bimodal biometric verification system

  • Authors:

  • Andrew Teoh Beng Jin;Salina Abdul Samad;Aini Hussain

  • Affiliations:

  • Faculty of Information Science and Technology, Multimedia University, Melaka, Malaysia;Department of Electrical, Electronic and System Engineering, Universiti Kebangsaan Malaysia, Bangi, Selangor, Malaysia;Department of Electrical, Electronic and System Engineering, Universiti Kebangsaan Malaysia, Bangi, Selangor, Malaysia

  • Venue:
  • AVBPA'03 Proceedings of the 4th international conference on Audio- and video-based biometric person authentication
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

A bimodal biometric verification system based on facial and vocal biometric modules is described in this paper. The system under consideration is built in parallel where each matching score reported by two classifiers are fused by using theoretic evidence k-NN (tekNN) based on Dempster-Safer (D-S) theory. In this technique, each nearest neighbour of a pattern to be classified is regarded as an item of evidence supporting certain hypotheses concerning the pattern class membership. Unlike statistical based fusion approaches, tekNN based on D-S theory is able to represent uncertainties and lack of knowledge. Therefore, the usage of tekNN leads to a ternary decision scheme, {accept, reject, inconclusive} which provides a more secure protection. From experimental results, the speech and facial biometric modules perform equally well, giving 93.5% and 94.0% verification rates, respectively. A 99.86% recognition rate is obtained when the two modules are fused. In addition, an 'unbalanced' case is been created to investigate the robustness of technique.