Comparative study of SVM & KNN for signature verification

Abstract

Signature is a unique identification of an individual and is a traditional means of human identification. It is very useful in legal activities in the physical absence of the person. Signature Verification is mostly used in forensic document analysis. In this paper, a comparative study is done on Support Vector Machine (SVM) & k-nearest neighbor algorithm (KNN) to find the accuracy of the algorithms. Genuine and Forged signature is taken to create training and testing dataset. Signature pixel values are stored in the dataset to train the algorithm. In this paper we are implementing a study between both the algorithm to find the advantages and disadvantages. Confusion matrix and ROC Curve is used to compare both algorithms accuracy and performance.

Subject Classification:

• Primary 62J05
• Secondary 32S22

Keywords:

• k-nearest neighbors’ algorithm (KNN)
• Signature Verification
• Biometrics
• Forensic signature examination
• Support Vector Machine (SVM)