Fingerprint is the generic structure in the form of a very detailed pattern and a sign that inherent in human beings. Many biometric systems using fingerprint as input data, because the nature of each individual is different although identical twins and do not change unless got a accident. The method used in this research is image segmentation using Otsu thresholding algorithm, feature extraction using Local Binary Pattern (LBP) algorithm and the learning method using Learning Vector Quantization (LVQ) algorithm. The used data is grayscale fingerprint image with 200x300 pixel and *.jpg extension format. The fingerprint image is composed of 25 people, each person has 6 training data and 2 test data. Experiment of training data and test data conducted for four systems, namely the system with characteristics of LBP = 8, 64, 128 and 256 and their respective uses 2 pieces of data set where data set 1 amounted to 15 people and data set 2 amounted to 25 people. The fourth experiment results show that the system is a system with a number of LBP characteristics = 128 is a system with the best combination of high system accuracy and fast learning time.