Evaluation of confidence measures for on-line handwriting recognition

Anja Brakensiek, Andreas Kosmala, Gerhard Rigoll

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

In this paper a writer-independent on-line handwriting recognition system is described comparing the effectiveness of several confidence measures. Our recognition system for single German words is based on Hidden Markov Models (HMMs) using a dictionary. We compare the ratio of rejected words to misrecognized words using four different confidence measures: One depends on the frame-normalized likelihood, the second on a garbage model, the third on a two-best list and the fourth on an unconstrained character recognition. The rating of recognition results is necessary for an unsupervised retraining or adaptation of recognition systems as well as for a user friendly human-computer interaction avoiding excessive call backs.

Original languageEnglish
Title of host publicationPattern Recognition - 24th DAGM Symposium, Proceedings
EditorsLuc Van Gool, Luc Van Gool, Luc Van Gool
PublisherSpringer Verlag
Pages507-514
Number of pages8
ISBN (Print)354044209X, 9783540442097
DOIs
StatePublished - 2002
Event24th Symposium of the German Pattern Recognition Association, DAGM 2002 - Zurich, Switzerland
Duration: 16 Sep 200218 Sep 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2449 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference24th Symposium of the German Pattern Recognition Association, DAGM 2002
Country/TerritorySwitzerland
CityZurich
Period16/09/0218/09/02

Fingerprint

Dive into the research topics of 'Evaluation of confidence measures for on-line handwriting recognition'. Together they form a unique fingerprint.

Cite this