Deformations in handwritten characters have class-dependent tendencies. For example, characters of class "A" are often deformed by global slant transformation and never deformed to be similar to "R". In this paper, the extraction and the utilization of such tendencies called eigen-deformations are investigated for better performance of elastic matching based recognition systems. The eigen-deformations are extracted by the principal component analysis of actual deformations automatically collected by elastic matching. From experimental results it was shown that the extracted eigen-deformations represent typical deformations of each class. It was also shown that the recognition performance can be improved significantly by using the eigen-deformations in detecting overfitting, which often results in misrecognition.
|Number of pages||4|
|Publication status||Published - 2002|
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering
- Computer Vision and Pattern Recognition
- Hardware and Architecture