Research on Feature Extraction Method for Handwritten Chinese Character Recognition Based on Kernel Independent Component Analysis

摘要

Feature extraction is very difficult for handwritten Chinese character because of large Chinese characters set, complex structure and very large shape variations. The recognition rate by currently used feature extraction methods and classifiers is far from the requirements of the people. For this problem, this study proposes a new feature extraction method for handwritten Chinese character recognition based on Kernel Independent Component Analysis (KICA). Firstly, we extract independent basis images of handwritten Chinese character image and the projection vector by using KICA algorithm and then obtain the feature vector. The scheme takes full advantage of good extraction local features capability of ICA and powerful computational capability of KICA. The experiments show that the feature extraction method based on KICA is superior to that of gradient-based about the recognition rate and outperforms that of ICA about the time for feature extraction.