Ink width independent global features for writer verification

The present work reports the results of writer verification under different writing conditions and different ink widths. It is observed that the style of a writer and the type of writing instrument used greatly affects the handwriting. The objective of our work is to verify a writer under the changing conditions. To improve verification rate, the directional information is captured and the histogram of chain code and differential chain code are used as features. The discriminating character used is `f´. The system is tested for 30 writers with 15 samples each with different writing instruments. Total of 270 samples are used to train the system and the system is tested for remaining 180 samples. The system is tested using chain code, first order differential chain code, second order differential chain code and curvature feature. Chain code is found to outperform the other methods with 2.76 % as the average error rate (AER).