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-stream HMM for EMG-based speech recognition.

H Manabe1, Z Zhang

  • 1Multimidia Laboratories, NTT DoCoMo, Inc., Kanagawa, Japan. manabe@mml.yrp.nttdocomo.co.jp

Conference Proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference
|February 3, 2007
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

Detection of Japanese eel endothelial cells-infecting virus (JEECV) in the Japanese eel Anguilla japonica (Temminck & Schlegel), living in natural habitats.

Journal of fish diseases·2014
Same author

Embolisation of Ruptured Cerebral Aneurysms with Interlocking Detachable Coils in Acute Stage. Acute IDC Embolisation for Ruptured Aneurysms.

Interventional neuroradiology : journal of peritherapeutic neuroradiology, surgical procedures and related neurosciences·2010
Same author

Long-term results of ruptured cerebral aneurysms embolized in acute stage with interlocking detachable coils.

Interventional neuroradiology : journal of peritherapeutic neuroradiology, surgical procedures and related neurosciences·2010
Same author

A case of partially thrombosed giant basilar aneurysm successfully treated by coil embolization followed by proximal clipping.

Interventional neuroradiology : journal of peritherapeutic neuroradiology, surgical procedures and related neurosciences·2010
Same author

Coil embolization for ruptured dissection on the vertebral artery distal to the origin of posterior inferior cerebellar artery.

Interventional neuroradiology : journal of peritherapeutic neuroradiology, surgical procedures and related neurosciences·2010
Same author

Serum high-density and low-density lipoprotein cholesterol is associated with abdominal aortic aneurysm presence: a systematic review and meta-analysis.

International angiology : a journal of the International Union of Angiology·2010

This study introduces an improved electromyography (EMG)-based speech recognition system using multi-stream hidden Markov models (HMMs). Optimizing stream weights enhanced accuracy by 4.0%, reducing errors by 12.8% for silent speech recognition.

Area of Science:

  • Biomedical Engineering
  • Signal Processing
  • Human-Computer Interaction

Background:

  • Electromyography (EMG) signals offer a non-invasive method for speech recognition by capturing muscle activity.
  • Existing EMG-based speech recognition systems can be limited by accuracy and the need for complex feature extraction.
  • Integrating established speech recognition technologies can potentially enhance EMG-based systems.

Purpose of the Study:

  • To propose and evaluate a novel technique for improving EMG-based speech recognition accuracy.
  • To investigate the effectiveness of multi-stream hidden Markov models (HMMs) for classifying facial EMG signals.
  • To enhance speech recognition capabilities for silent speech (mouth movements only).

Main Methods:

  • Facial EMG signals were collected from ten subjects uttering ten Japanese isolated digits using a 3-channel system.

Related Experiment Videos

  • A multi-stream HMM architecture was employed, with each EMG channel serving as a distinct stream.
  • Feature extraction focused on static parameters and their delta components, with optimized stream weighting.

    • Main Results:

    • The delta component of the static parameter was identified as a key feature for improving recognition accuracy.
    • Individual optimization of stream weights resulted in a 4.0% increase in recognition accuracy compared to equal weighting.
    • This optimization led to a significant 12.8% reduction in the overall error rate.

    Conclusions:

    • Multi-stream HMMs are effective for classifying EMG signals in the context of speech recognition.
    • Optimized stream weighting in multi-stream HMMs significantly enhances EMG-based speech recognition performance.
    • The proposed technique offers a promising approach for developing accurate silent speech recognition systems.