A Chinese Minority Script Recognition Method Based on Wavelet Feature and Modified KNN

Abstract

In recent years, K-Nearest Neighbor (KNN) has demonstrated excellent performance in a variety of pattern recognition problems. In this paper, we proposed modified KNN for the Chinese minority scripts classification, using wavelet energy distribution and wavelet energy proportions features generated from the discrete wavelet multiresolution decomposition. The experimental results show that the method achieves high accuracy in testing dataset that consists of Chinese, English and Chinese minority scripts such as Tibetan, Tai Lue, Naxi Pictographs, Uighur, Tai Le, Yi. The results also show that this method is feasible and reasonable, and the recognition rate is up to 96%.