Computation of Probabilities for an Island-Driven Parser
IEEE Transactions on Pattern Analysis and Machine Intelligence
IEEE Transactions on Pattern Analysis and Machine Intelligence
Extraction of key letters for cursive script recognition
Pattern Recognition Letters
Off-Line Handwritten Word Recognition Using a Hidden Markov Model Type Stochastic Network
IEEE Transactions on Pattern Analysis and Machine Intelligence
Building a Perception Based Model for Reading Cursive Script
ICDAR '95 Proceedings of the Third International Conference on Document Analysis and Recognition (Volume 2) - Volume 2
Strategies in character segmentation: a survey
ICDAR '95 Proceedings of the Third International Conference on Document Analysis and Recognition (Volume 2) - Volume 2
Recognizer characterisation for combining handwriting recognition results at word level
ICDAR '95 Proceedings of the Third International Conference on Document Analysis and Recognition (Volume 1) - Volume 1
A hybrid radial basis function network/hidden Markov model handwritten word recognition system
ICDAR '95 Proceedings of the Third International Conference on Document Analysis and Recognition (Volume 1) - Volume 1
Building a large annotated corpus of English: the penn treebank
Computational Linguistics - Special issue on using large corpora: II
An HMM-Based Threshold Model Approach for Gesture Recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
Hi-index | 0.00 |
For on-line handwriting recognition, a hybrid approach that combinesthe discrimination power of neural networks with the temporalstructure of hidden Markov models is presented. Initially, allplausible letter components of an input pattern are detected by usinga letter spotting technique based on hidden Markov models. A wordhypothesis lattice is generated as a result of the letter spotting.All letter hypotheses in the lattice are evaluated by a neuralnetwork character recognizer in order to reinforce letterdiscrimination power. Then, as a new technique, an island-drivenlattice search algorithm is performed to find the optimal path on theword hypothesis lattice which corresponds to the most probable wordamong the dictionary words. The results of this experiment suggestthat the proposed framework works effectively in recognizing Englishcursive words. In a word recognition test, on average 88.5% wordaccuracy was obtained.