Abstract
Gaussian Mixture Modeling, usually known by its acronym GMM, is a statistical pattern recognition technique usually applied to speaker identification. This technique will be proved to be used with morphological features extracted from images. Therefore, this technique will be applied to hand geometry biometrics in order to improve the results obtained with other verification techniques. With the results that will be shown, this biometric technique, which is considered a low-medium security one, can be applied to high security environments, due to its error rates below 5%. The features to enter the GMM algorithm will be extracted from a color image of the hand (both the top and the side view of it). This image will be processed and its edged detected before a morphological analysis is performed, obtaining a feature vector as small as 12 bytes. In this paper, after a short introduction and a general overview of the GMMs the whole process will be explained.