Research on Character Recognition Technology of New Tai Lue Based on Gabor Feature and SVM

Abstract

New Tai Lue script which is also known as Simplified Tai Lue, is an alphabet used to write the Tai Lü language used in Xishuangbanna of Yunnan Province in China. In this paper, we present a handwritten character recognition method of New Tai Lue by Support vector machine (SVM) based classifier on Gabor Features. First, some preprocessing approaches such as superfluous points removing, normalization, smoothing, resampling are applied to preprocess the handwritten character images of New Tai Lue. Second, a filter bank consisting of two-dimensional Gabor filters with various scales and orientations is created to extract features from these images. After that, the result images are processed by PCA to do dimension reduction. Finally, SVM is applied as the classifier to perform recognition task. Experimental results conducted on a database of 16200 characters show that the proposed method are effective.