Signature is one of the biometric humans that used widely. This system aims to recognize the signature through feature extraction and image classification method signature with Support Vector Machine (SVM). Research databases used 15 samples signatures images from students of Informatics Engineering UNIB with size 300 x 300 pixels. The system is built in the Java programming language with NetBeans IDE 8.0. The system is divided into 3 stages: preprocessing, feature extraction, and SVM classification. Preprocessing stages are binerization, noise Removal, thinning, cropping, and resizing. Feature extraction stage using Image Centroid Zone (ICZ) and Zone Centroid Zone (ZCZ) methods. Furthermore, the results of feature vectors ICZ and ZCZ be input training SVM classification. This research results shows that: (1) the greater the size of the zone, the higher the identification accuracy; (2) the smaller polynomial degree, the higher the signatures identification accuracy; (3) The best performance is obtained for 5x4 size zone and 2 degree polynomial with 97.33% signature identification accuracy.