Fingerprint and speaker recognition using artificial neural networks

Keywords: Neural Networks, Speaker Recognition, Fingerprint Recognition Biometric, is a automatical definition that based on physiological and behavioral properties of the people. It is more better to diffirentiate authorized and cribbler person than methods like traditional password and personal definition numbers. In this study, required parameters were obtained by analyzing the biometric features fingerprint and speaking signals. Ending points and bifurcation points which are the most important ones of feature points named as Galton Characteristics were determined from fingerprint samples, and the distances of these points to the core point were taken as references. Required parameters were obtained from speaking signals by MFCC (Mel Frequency Cepstrum Coefficients) method. Identification was made by a three layer neural network structure which consists of input, hidden and output layers and using the backpropagation algorithm "trainscg" with the parameters obtained. After the network was trained, when it was tested with the fingerprint and speaking signal samples and 100% of the results were right. When the test samples were used, these values were 77% for the fingerprints, and 86% for the speech samples.