
	Journal of Applied Sciences
	Year: 2012 \| Volume: 12 \| Issue: 12 \| Page No.: 1282-1289 DOI: 10.3923/jas.2012.1282.1289

Performances of Qualitative Fusion Scheme for Multi-biometric Speaker Verification Systems in Noisy Condition

Lydia Abdul Hamid and Dzati Athiar Ramli

Abstract: Fusion of multiple modalities becomes a good strategy to improve the performance in biometric speaker verification as an audio based system accuracy decreases severely under noisy condition. However, for simple sum rule fusion scheme, this approach is only helpful if both systems have equal performances because it considers both models equally regardless of its conditions. As a consequence, a weighted sum rule is then experimented. Instead of varying the weight, this method calculates the sum of Equal Error Rate (EER) percentages produced from each modality for fusion weight computation. In this study, the information in term of Mel Frequency Cepstral Coefficient (MFCC) of speech signal is extracted while Region of Interest (ROI) of lip images has been used as a second modality for the multi-modal systems. The Support Vector Machine (SVM) classifier is executed for the verification system. From the experiment results, EER performances using simple sum rule and weighted sum rule at 35 dB SNR of speech signal and 0.9 noise density of visual signal are observed as 0.1548 and 0.1487%, respectively.

PDF Fulltext XML References Citation Report Citation

RELATED ARTICLES:

•	ARM Based 3-in-1 Device for People with Disabilities

•	Improving Speaker Verification in Noisy Environments using Adaptive Filtering and Hybrid Classification Technique

•	Modular Arithmetic and Wavelets for Speaker Verification

•	Speaker Recognition Based on Mathematical Morphology

•	A UMACE Filter Approach to Lipreading in Biometric Authentication System

How to cite this article:

Lydia Abdul Hamid and Dzati Athiar Ramli, 2012. Performances of Qualitative Fusion Scheme for Multi-biometric Speaker Verification Systems in Noisy Condition. Journal of Applied Sciences, 12: 1282-1289.

DOI: 10.3923/jas.2012.1282.1289

URL: https://scialert.net/abstract/?doi=jas.2012.1282.1289

COMMENT ON THIS PAPER

Navigation


		Online First
		Current Issue
		Previous Issues
		Editorial Board
		Submit a Manuscript
		Guide to Authors
		Subscribe to E-alerts

Indexed In


		AGRIS
		ASCI-Database
		Asian Digital Library
		Chemical Abstract Services
		Google Scholar

Science Alert