A novel multistage classification and Wavelet based kernel generation for handwritten Marathi compound character recognition

This paper presents a novel approach for recognition of unconstrained handwritten Marathi compound characters. The recognition is carried out using multistage feature extraction and classification scheme. The initial stages of feature extraction are based upon the structural features and the classification of the characters is done according to their parameters. The final stage of feature extraction employs generation of kernels using Wavelet transform. A single level Wavelet decomposition is used to generate the approximation coefficients. These coefficients are stored as kernels for matching. A modified wavelet based kernel generation method is also implemented. The recognition is done by template matching in both the cases. The results are analyzed using both the kernel generation techniques for varying resize factors. The recognition rate achieved from the proposed method is 95.89% and 96.00% for 16×16 and 32×32 resize factors respectively with wavelet based kernels and 96.41% and 97.94% for 16×16 and 32×32 resize factors respectively with modified wavelet based kernels.