Signatures continue to be an important biometric for authenticating the identity of human beings. This paper presents an effective method to perform off-line signature verification using unique structural features extracted from the signature's contour. A novel combination of the Modified Direction Feature (MDF) and additional distinguishing features such as the centroid, surface area, length and skew are used for classification. A Resilient Backpropagation (RBP) neural network and a Radial Basis Function (RBF) network were compared in terms of verification accuracy. Using a publicly available database of 2106 signatures (936 genuine and 1170 forgeries), verification rates of 91.21% and 88.0% were obtained using RBP and RBF respectively.