Handwriting digit recognition using local binary pattern variance and K-Nearest Neighbor classification

This paper present about handwriting digit recognition system using Local Binary Pattern as feature extraction method and K-Nearest Neighbor as classification algorithm. This system is implemented on handwriting image on the C1 form used by General Elections Commision in Indonesia to facilitate the committee member to input the election result to the database. This system consist of several steps, such as data collecting, pre-processing, feature extraction and classification. The testing result on this system shows that the Local Binary Pattern Variance method can recognize handwriting digit character on MNIST dataset with accuracy level 89,81% using the best parameter value radius=4,256 and 64-bin histogram, 9 region division on the image, and 10 nearest neighbor on K-NN algorithm. When tested using the data from C1 form, the system accuration value is 70,91%.