摘要:
Neural Network is widely used in pattern recognition and classification. The proposed system aims to implement an offline signature verification system using neural network. Before extracting the features, rotated signature image is normalized to achieve the corrected signature image. Without retrieving of the whole feature of the signature image, signature image is segmented to achieve 64 sub-images based on center of gravity. Three features are extracted from each sub-image block. These features are used as an input to neural network. In verification, test signature's features are used to a trained neural network to verify the signature as an accepted or rejected signature.
展开