Multimodal biometric system based on face and hand images taken by a cell phone.

摘要

One of the methods to improve the recognition rate of humans is multimodal biometrics, which is based on more than one physiological or behavioral characteristics to identify an individual. Multimodal biometrics improves not only the performance, but also nonuniversality and spoofing that are commonly encountered in unibiometric systems. In this thesis, we built a multibiometric system that works on face and hand images taken by a camera built into a cell phone. The multimodal fusion is done at the feature extraction level. The nine facial models are built according to the number of features/points extracted from the face. Active shape models method is applied in order to find the concatenated string of facial points in the eyes, nose, and mouth areas. The face feature vector is constructed by applying Gabor filter to the image and extracting the key points found by an active shape model. The hand feature vector contains nine geometric measurements, including heights and widths of four fingers, and the width of the palm. Support vector machine is used as a classifier for a multimodal approach. One SVM machine is built for each person in the database to distinguish that person from the others. The database contains 113 individuals. As the experiments show, the best accuracy of up to 99.82% has been achieved for the multibiometric model combining 8 eye points, 4 nose points, and 9 hand features.