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

Multi-aspect target discrimination using hidden Markov models and neural networks.

Marc Robinson1, Mahmood R Azimi-Sadjadi, Jaime Salazar

  • 1Department of Electrical and Computer Engineering, Colorado State University, Fort Collins, CO 80523, USA.

IEEE Transactions on Neural Networks
|March 25, 2005
PubMed
Summary
This summary is machine-generated.

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

Impact of carboplatin desensitization therapy on progression-free survival in gynecologic cancers.

Gynecologic oncology reports·2026
Same author

A Rare Case of Recurrent Sarcoma Presenting as an Anterior Mediastinal Mass: Case Report.

Case reports in oncology·2026
Same author

Association of social vulnerability with access to treatment in ovarian cancer patients.

Gynecologic oncology reports·2025
Same author

Characterization of Dermatologic Disease and Challenges in Care of Incarcerated Patients.

Journal of correctional health care : the official journal of the National Commission on Correctional Health Care·2025
Same author

Infectious Disease Education in a Carceral Setting: Insights from a Jail-Based Medical Student-Led Health Literacy Initiative.

Journal of correctional health care : the official journal of the National Commission on Correctional Health Care·2024
Same author

Voices Behind Bars: Ethical, Legal, and Practical Considerations for Surrogates Who Are Incarcerated.

Annals of internal medicine·2023

A novel hybrid system combines multilayer perception (MLP) networks with hidden Markov models (HMMs) for enhanced multi-aspect pattern classification. This approach improves underwater target identification using sonar data.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Pattern Recognition

Background:

  • Multi-aspect pattern classification is crucial for complex data analysis.
  • Hidden Markov Models (HMMs) are effective for sequential data but can be limited in emission probability estimation.
  • Multilayer Perception (MLP) networks excel at function approximation and probability estimation.

Purpose of the Study:

  • To introduce a novel hybrid system integrating HMMs and MLPs for multi-aspect pattern classification.
  • To enhance the accuracy of underwater target classification using sonar data.
  • To develop an optimized method for estimating HMM parameters.

Main Methods:

  • A hybrid model combining HMMs for state transitions and MLPs for emission probabilities.

Related Experiment Videos

  • A new batch gradient descent algorithm for optimal estimation of transition and emission probabilities.
  • A predictive method utilizing previous states to improve transition probability computation, exploiting aspect correlations.

    • Main Results:

    • The hybrid HMM-MLP system demonstrated effective multi-aspect pattern classification.
    • The batch gradient descent method achieved optimal estimation of model parameters.
    • The prediction method enhanced transition probability calculations by leveraging sequential data correlations.

    Conclusions:

    • The proposed hybrid HMM-MLP system offers a powerful approach for multi-aspect pattern classification.
    • The optimized parameter estimation and predictive methods contribute to improved classification accuracy.
    • The system's efficacy is validated on a challenging underwater target classification task using real-world sonar data.