Minimum classification error training for online handwritten word recognition

被引：3

作者：

Biem, A ^{[1
]}

机构：

[1] IBM Corp, Thomas J Watson Res Ctr, Yorktown Hts, NY 10598 USA

来源：

EIGHTH INTERNATIONAL WORKSHOP ON FRONTIERS IN HANDWRITING RECOGNITION: PROCEEDINGS | 2002年

关键词：

D O I：

10.1109/IWFHR.2002.1030885

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We describe an application of the Minimum Classification Error (MCE) training criterion to online unconstrained-style word recognition. The described system uses allograph-HMMs to handle writer variability. The result, on vocabularies of 5k to 10k, shows that MCE training achieves around 17% word error rate reduction when compared to the baseline Maximum Likelihood system.

引用

页码：61 / 66

页数：4

共 12 条

[1] Pattern recognition using discriminative feature extraction
Biem, A
Katagiri, S
Juang, BH
[J]. IEEE TRANSACTIONS ON SIGNAL PROCESSING, 1997, 45 (02) : 500 - 504
[2] BIEM A, 2001, P ICASSP 3
[3] BIEM A, 1997, THESIS U PARIS 6
[4] Chen W.-T., 2000, P 7 INT WORKSH FRONT, P393
[5] FAHLMAN SE, 1998, CMUCS88162
[6] HUO Q, 2001, P ICASSP, V3
[7] DISCRIMINATIVE LEARNING FOR MINIMUM ERROR CLASSIFICATION
JUANG, BH
KATAGIRI, S
[J]. IEEE TRANSACTIONS ON SIGNAL PROCESSING, 1992, 40 (12) : 3043 - 3054
[8] KATAGIRI S, 1991, P IEEE WORKSH NEUR N, P309
[9] A DECISION THEORETIC FORMULATION OF A TRAINING PROBLEM IN SPEECH RECOGNITION AND A COMPARISON OF TRAINING BY UNCONDITIONAL VERSUS CONDITIONAL MAXIMUM-LIKELIHOOD
NADAS, A
[J]. IEEE TRANSACTIONS ON ACOUSTICS SPEECH AND SIGNAL PROCESSING, 1983, 31 (04): : 814 - 817
[10] NATHAN KS, 1995, P ICASSP 1995 IEEE I, V4, P2619

← 1 2 →