On-Line Handwriting Recognition Using Hidden Markov Models
نویسنده
چکیده
New global information-bearing features improved the modeling of individual letters, thus diminishing the error rate of an HMM-based on-line cursive handwriting recognition system. This system also demonstrated the ability to recognize on-line cursive handwriting in real time. The BYBLOS continuous speech recognition system, a hidden Markov model (HMM) based recognition system, is applied to on-line cursive handwriting recognition. With six original features, delta x, delta y, writing angle, delta writing angle, PenUp/PenDown bit, and sgn(x-max(x)), the baseline system obtained a word error rate of 13.8% in a 25K-word lexicon, 86-character set, writer-independent task. Four new groups of features, a vertical height feature, a space feature, hat stroke features, and substroke features, were implemented to improve the characterization of vertical height, inter-word space, and other global information. With the new features, the system obtained a word error rate of 9.1%, a 34% reduction in error. Additionally, the space feature and the substroke features each reduced the word error rate approximately 15%. In addition, we demonstrated real-time, large vocabulary, writer-independent, online cursive handwriting recognition without sacrificing much recognition accuracy of the baseline system by implementing minor modifications to the baseline handwriting recognition system. The details of the on-line cursive handwriting recognition system, the feature experiments, and the real-time demonstration system are presented. Thesis Supervisors: Dr. John Makhoul and Dr. Victor W. Zue Titles: Chief Scientist, Speech and Language Department, BBN Corp. and Senior Research Scientist, Dept. of EECS, MIT
منابع مشابه
Improved Modeling in Handwriting Recognition
In this work a script independent handwriting recognition system is proposed which is derived from the RWTH-ASR hidden Markov model (HMM) based speech recognizer. Most problems occurring in handwriting recognition (HWR) are induced by large variations within the written text. In particular, different handwriting styles such as cursive writing or long drawn-out strokes are difficult to model. Co...
متن کاملImproved On-Line Handwriting Recognition Using Context Dependent Hidden Markov Models
This paper presents the introduction of context dependent Hidden Markov Models for cursive, uncon-strained handwriting recognition with large vocabularies. Since context dependent models were successfully introduced to speech recognition ((1], 2], 3]), it seems obvious, that the use of trigraphs could also lead to improved on-line handwriting recognition systems 4]. In analogy to triphones in s...
متن کاملAn Empirical Evaluation of Off-line Arabic Handwriting And Printed Characters Recognition System
Handwriting recognition is a challenging task for many real-world applications such as document authentication, form processing, historical documents. This paper focuses on the comparative study on off-line handwriting recognition system and Printed Characters by taking Arabic handwriting. The off-line Handwriting Recognition methods for Arabic words which being often used among then across the...
متن کاملUnsupervised Estimation of Writing Style Models for Improved Unconstrained Off-line Handwriting Recognition
The performance of writer-independent unconstrained handwriting recognition is severely affected by variations in writing style. In a segmentation-free approach based on Hidden-Markov models we, therefore, use multiple recognition models specialized to specific writing styles in order to improve recognition performance. As the explicit definition of writing styles is not obvious we propose an u...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2002