Large scale hand-written character recognition system using subspace method

The subspace method proposed by Watanabe offers the basic concept of subspace construction, but the issue of how to use the limited samples to construct effective subspace to avoid the problem of the subspace inclining toward mean vectors remains unresolved. To cope with this problem, the authors have proposed the combination method (CM), which constructs the subspace from several groups including different number of samples divided from the whole training samples. The CM obtained a high recognition rate of 97.76% with respect to ETL9B, the largest database of hand-written characters in Japan. Next, the issues of how to improve the recognition accuracy and how to accelerate the recognition speed are dealt with. In this paper, we propose a new method called the uniform division method (UDM), which uses the uniformly divided training samples to construct a subspace. Compared to the CM given earlier, the UDM is very simple and effective enough to improve the accuracy of recognition. The UDM algorithm and the experiments with ETL9B are described.