Classifications of Signatures by Radial Basis Neural Network

The personal signature can be considered one of the most common behavioral biometrics. In this study, signatures are classified according to their specifications. The statistical calculation is considered for the specifications of each signature. Then, a radial basis neural network (RBNN) is adapted to apply multiple classifications for the employed signatures. A big number of signatures are utilized; they are obtained from the database called biometric ideal test (BIT). The total number of collected signatures is equally divided between the testing and training phases, where it is partitioned into 50% for the training and 50% for the testing. The proposed technique could achieve attractive performance, where each of the mean square error (MSE) and mean absolute error (MAE) attained a small value of 0.028. In addition, the proposed approach using the RBNN is compared with the different neural networks of the state-of-the-art techniques in order to demonstrate that the outcomes are acceptable and successful.