Recently, several new methodologies based on statistical schemes have appeared that produce successful results in the field of pattern recognition. This paper presents two of them, neural network classifiers and hidden Markov models (HMMs), and further proposes a hybrid of both. The hybrid method exploits the discrimination capability of neural network classifiers while using HMM formalism to capture the dynamics of input patterns. The presented methods are applied to the recognition of on-line handwritten characters. Experimental results with a large set of multi-writer data confirm the usefulness of more sophisticated methodologies such as multistage schemes and multiple network schemes as well as the hybrid method.
|Number of pages||9|
|Journal||Computer Standards and Interfaces|
|Publication status||Published - 1994 Jul|
Bibliographical noteFunding Information:
* Email: sbcho(~'hip.atr.co.jp, email@example.com * This work was supported in part by a grant from the Korea Science and Engineering Foundation (KOSEF) and Center for Artificial Intelligence Research (CAIR), the Engineering Research Center (ERC) of Excellence Program.
All Science Journal Classification (ASJC) codes
- Hardware and Architecture