Recognition of printed Thai characters using a matching method

As there are many similar characters in printed Thai, and in order to get a high recognition rate, the recognition system is separated into two stages. In the rough classification stage, fine features and noise are ignored by blurring. The blurred characters are separated into some cluster domains. The clustering criterion used is based on selection of the patterns by measuring the similarity coefficient. The Karhunen-Loeve expansion is applied to get a standard pattern of each category. In the fine classification stage, subpattern matching is used to discriminate between the characters.