<![CDATA[Signature is a personal attribute that has long been widely accepted as a tool for verification of personal identity. But signatures are also easy to fake to be misused. To avoid this, a system is created to verify signatures. This system uses the image of the signature captured by the camera as an input triggered by the push button, Raspberry Pi as a digital image processing unit, and LCD 16x2 as the system output. This study uses the Histogram of Oriented Gradients (HOG) feature descriptor with precedence of image preprocessing. The output of the HOG method is a feature vector that represents the signature characteristics of the image, this feature vector which will be classified with the Support Vector Machine (SVM) classifier for data training and data prediction. There are two main parts of system software, the training data section, and the testing data section for signature verification. The implementation results obtained that the system can verify signatures with an accuracy of 87.33%. System requires 1.45 seconds in average to train data on each signatory name and for the verification process, the average system takes 0.238 seconds for the genuine signature and 0.242 seconds for forgery signatures.]]>
Copyrights © 2019
