A Database for Handwritten Text Recognition Research
IEEE Transactions on Pattern Analysis and Machine Intelligence
Determination of the Number of Writing Variants with an HMM based Cursive Word Recognition System
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
| Hi-index | 0.00 |
On the basis of a well accepted, HMM-based cursivescript recognition system, an algorithm which automaticallyadapts the length of the models representing the letterwriting variants is proposed.An average improvement inrecognition performance of about 2.72 percent could be obtained.Two initialization methods for the algorithm havebeen tested, which show quite different behaviors; bothprove to be useful in different application areas.To geta deeper insight into the functioning of the algorithm amethod for the visualization of letter HMMs is developed.It shows the plausibility of most results, but also the limitationsof the proposed method.However, these are mostlydue to given restrictions of the training and recognitionmethod of the underlying system.