Abstract

There are many biometric authentication systems that have used different human traits either physiological or behavioural which are liable to forgery. Thus, the human tongue promises to deliver a level of uniqueness that other biometrics cannot match because it is immune to forgery. However, literature has revealed that a single feature extraction technique was not sufficient for effective and efficient treatment of biometrics because of lapses imposed by individual technique. This paperpresents the development of a tongue recognition system using Gabor filter and particle swarm optimization. One hundred and twenty tongue images of human tongues of different individuals, six tongue images with different expressions were captured each per person with a digital camera. The tongue images were pre-processed through extraction of region of interest and normalization usingbinary mask and histogram equalization; Gabor filter was used for feature extraction and particle swarm optimization (PSO) was used to select salient features.Some samples tongue images were matched with the stored extracted tongue images using Euclidean distance classifier. The results show that at threshold value of 0.50, G-PSO has a sensitivity, specificity and accuracy rate of 100.00%, 94.44% and 98.08% respectively compared to Gabor which has 94.12%, 88.89% and 92.31% respectively and PSO which has 97.06%, 88.89% and 94.23% respectively.

Authors and Affiliations