Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Improving offline handwritten text recognition with hybrid HMM/ANN models.

Salvador España-Boquera1, Maria Jose Castro-Bleda, Jorge Gorbe-Moya

  • 1Departamento de Sistemas Informáticos y Computación, Universitat Politècnica de València, Camino de Vera s/n, 46022 Valencia, Spain. sespana@dsic.upv.es

IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 18, 2010
PubMed
Summary

This study introduces hybrid Hidden Markov Model (HMM)/Artificial Neural Network (ANN) models for improved unconstrained offline handwritten text recognition. New preprocessing techniques enhance accuracy by normalizing text size and correcting slope and slant.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Leveraging Transformers-based models and linked data for deep phenotyping in radiology.

Computer methods and programs in biomedicine·2025
Same author

First-Episode Psychotic Patients Showed Longitudinal Brain Changes Using fMRI With an Emotional Auditory Paradigm.

Frontiers in psychiatry·2021
Same author

Fallback Variable History NNLMs: Efficient NNLMs by precomputation and stochastic training.

PloS one·2018
Same author

Efficient Embedded Decoding of Neural Network Language Models in a Machine Translation System.

International journal of neural systems·2018
Same author

Supervised filters for EEG signal in naturally occurring epilepsy forecasting.

PloS one·2017
Same author

Crowdsourcing reproducible seizure forecasting in human and canine epilepsy.

Brain : a journal of neurology·2016

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Pattern Recognition

Background:

  • Handwritten text recognition is challenging due to variations in writing styles.
  • Existing methods often struggle with unconstrained and offline data.
  • Preprocessing is crucial for improving recognition accuracy.

Purpose of the Study:

  • To propose and evaluate hybrid Hidden Markov Model (HMM)/Artificial Neural Network (ANN) models for unconstrained offline handwritten text recognition.
  • To introduce novel supervised learning techniques for text image preprocessing.
  • To achieve state-of-the-art recognition rates on benchmark datasets.

Main Methods:

  • Hybrid HMM/ANN models combining Markov chains for structure and Multilayer Perceptrons for emission probabilities.

Related Experiment Videos

  • Supervised learning methods for slope correction and size normalization using text contour extrema.
  • Non-uniform slant removal using Artificial Neural Networks.

    • Main Results:

    • The proposed hybrid models demonstrate high performance on offline handwritten text lines.
    • Preprocessing techniques effectively remove slope, slant, and normalize text size.
    • Achieved recognition rates are competitive with, and among the best reported in, the literature.

    Conclusions:

    • Hybrid HMM/ANN models offer a powerful approach for handwritten text recognition.
    • Advanced preprocessing significantly boosts recognition accuracy.
    • The developed methods represent a substantial advancement in the field.