<![CDATA[Signature have been widely accepted by people as one of many tools that used to verify a person's identity. Signature verification is highly needed in order to avoid crimes regarding signature's validity. The process of signature image verification uses feature extraction based on Pyramid Histogram of Oriented Gradient (PHOG) for extracting feature from global to local image area that used for the next process, classification using K-Nearest Neighbor method. There are some parameters that can affect the feature extraction of PHOG and K-NN as classification method such as number of bin, level, range of angles, and K. As for the additional parameters, namely the amount of training data that affect the overall results of the classification used. Feature extraction and classification by the method with the best parameter values and training data used produces the highest accuracy of 99.5% on Indonesian original signature data and 98.5% on the data of the Persian original signatures. While the forgery signatures data produces accuray only as much as 56% on data from Indonesia and 35,5% on data from Persian. Results from tests show that the algorithm is not good enough for distinguishing forgery signature that has high similarity with genuine signature even it is works well for recognizing genuine signature.]]>
Copyrights © 2017
