An implementation of Gabor based adaptive matching pursuit for multi-modal biometric registration

This research describes the design and implementation of a novel Gabor-based Adaptive Matching Pursuit Algorithm applied to voice data for Multi-modal Biometrics Registration. Combining the resultant voice feature hyperspace vector with the average absolute deviation feature vector created by [Rosa2005], we are able to fuse the two data types together at the feature level into a 940 dimensional multi-modal biometrics feature vector. Our multi-modal biometrics feature extraction system attempts to alleviate some of the weaknesses of single mode systems by introducing multiple modes of sensor data which when combined, make for a more robust system that is less prone to spoofing, mechanical failure and error caused by noise in single mode biometrics.