Discrete Meyer Wavelet Transform Features For online Hangul Script Recognition

Online hangul script recognition is important when writers input characters into computer and communication apparatus (such as PDA, Mobile Phone). In this study, a Wavelet Transform Features-based method for performance improvement of online handwritten hangul character recognition is proposed. The main idea is applying the Discrete Wavelet Transform (DWT) spectral analysis to the recognition of online hangul script. This method is based on the fact that online scripts offer space and time information. Locations of sample points belonging to a script give only space information and the order of occurrences of sample points provides time information. Given an online handwritten character sample, after a series of preprocessing, we obtain a 64×64 normalized online hangul handwritten script with the time information. The order of sample points can be the index of sequences. One sequence is the vertical coordinate of sample points. The second sequence is the horizontal coordinate of sample points. The third sequence is the product of the vertical coordinate and horizontal coordinate of sample points. The fourth sequence is the ratio between the vertical coordinate difference and horizontal coordinate difference of two sample points. The four sequences are combined as a vector whose size is 512. The vector is convoluted with the Meyer Wavelet and its dimension is reduced from 512 to 128 by Linear Discriminant Analysis (LDA) scheme. Modified Quadratic Discriminant Functions (MQDF) is utilized as the classifier for charter recognition. The Experiment results demonstrate that the method can improve the accuracy of character recognition.