Title :
Efficient utilization of variable duration information in HMM based HWR systems
Author :
Kundu, Amlan ; He, Yang ; Chen, Mou-Yen
Author_Institution :
US West Adv. Technol., Boulder, CO, USA
Abstract :
We describe an MD-HMM (model discriminant HMM) based handwritten word recognition (HWR) system (called NEHMM-nonergodic HMM) whose system parameters are derived from the parameters of the VDHMM (variable duration HMM) system described by Chen, Kundu and Srihari (see IEEE Trans. on Image Proc., vol.4, no.12, p.1675-88, 1995). The new HMM achieves better experimental results by a more efficient utilization of variable duration information. However, more often the problem is `reliable computation´ of duration probabilities given limited databases. A scheme (VSLHMM-variable sequence length HMM) has been presented to avoid the computation of duration probabilities altogether without sacrificing the performance gain of the VDHMM system
Keywords :
handwriting recognition; hidden Markov models; image segmentation; image sequences; probability; HMM based HWR systems; HWR system; MD-HMM; NEHMM; VDHMM system; databases; duration probabilities; experimental results; handwritten word recognition; image segmentation; image sequences; model discriminant HMM; nonergodic HMM; performance gain; reliable computation; system parameters; variable duration HMM; variable duration information; variable sequence length HMM; Character recognition; Data preprocessing; Dictionaries; Feature extraction; Helium; Hidden Markov models; Image databases; Image segmentation; Maximum likelihood estimation; Performance gain;
Conference_Titel :
Image Processing, 1997. Proceedings., International Conference on
Conference_Location :
Santa Barbara, CA
Print_ISBN :
0-8186-8183-7
DOI :
10.1109/ICIP.1997.632099
