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 Video

Updated: Jul 1, 2026

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

Sign language recognition by combining statistical DTW and independent classification.

Jeroen F Lichtenauer1, Emile A Hendriks, Marcel J T Reinders

  • 1Delft University of Technology, Faculty ofElectrical Engineering, Mathematics, and Computer Science, Informationand Communication Theory Group, Mekelweg 4, 2628 CD Delft, The Netherlands. j.lichtenauer@imperial.ac.uk

IEEE Transactions on Pattern Analysis and Machine Intelligence
|September 13, 2008
PubMed
Summary

Related Concept Videos

Classification of Signals01:30

Classification of Signals

In signal processing, signals are classified based on various characteristics: continuous-time versus discrete-time, periodic versus aperiodic, analog versus digital, and causal versus noncausal. Each category highlights distinct properties crucial for understanding and manipulating signals.
A continuous-time signal holds a value at every instant in time, representing information seamlessly. In contrast, a discrete-time signal holds values only at specific moments, often denoted as x(n), where...

You might also read

Related Articles

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

Sort by
Same author

Precision-guided immunomodulatory therapy in sepsis.

The Lancet. Respiratory medicine·2026
Same author

On the state of protein function prediction: a report on the fourth CAFA challenge.

bioRxiv : the preprint server for biology·2026
Same author

Effects of a Combined Dietary and Physical Activity Intervention on Bone Density, Lean Mass and Fat Mass in Adults: The GOTO Trial.

Journal of cachexia, sarcopenia and muscle·2026
Same author

Decoding exon inclusion in the human brain reveals more divergent splicing mechanisms in neurons than glia.

Genome biology·2026
Same author

Domain mapping of disease mutations reveals pathogenic SORL1 variants in Alzheimer's disease.

Molecular neurodegeneration·2025
Same author

All-atom protein sequence design using discrete diffusion models.

Journal of cheminformatics·2025

Separating time warping from classification improves recognition accuracy. New methods using Statistical Dynamic Time Warping (SDTW) for warping and novel classifiers for classification outperform existing approaches like Hidden Markov Models (HMM).

Area of Science:

  • Pattern Recognition
  • Machine Learning
  • Signal Processing

Background:

  • Hybrid approaches combining Dynamic Time Warping (DTW) or Hidden Markov Models (HMM) with discriminative classifiers are common for speech, handwriting, and sign language recognition.
  • Existing methods rely heavily on the likelihood models of DTW/HMM, which may present conflicting demands for time warping and classification.

Purpose of the Study:

  • To investigate the hypothesis that separating time warping and classification can overcome limitations of current hybrid methods.
  • To propose and evaluate novel methods for improved recognition accuracy by decoupling these two processes.

Main Methods:

  • Proposed a novel approach using Statistical DTW (SDTW) solely for time warping, followed by classification using distinct methods.
  • Introduced two new statistical classifiers, CDFD and Q-DFFM, utilizing discriminative features (DF).

Related Experiment Videos

Last Updated: Jul 1, 2026

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
08:25

Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment

Published on: May 7, 2019

  • Evaluated performance against HMM and standard SDTW, including experiments on model-combining strategies.

    • Main Results:

    • The proposed classifiers (CDFD, Q-DFFM) demonstrated superior performance compared to HMM and SDTW.
    • Combining likelihoods of multiple models degraded the performance of the proposed classifiers but improved HMM and SDTW.
    • A proof-of-concept showed that hybrid classification combining DFFM mappings and SDTW likelihoods significantly improved upon SDTW alone.

    Conclusions:

    • Separating time warping and classification is a beneficial strategy for recognition tasks.
    • The proposed SDTW-based warping with novel discriminative classifiers offers a promising alternative to traditional HMM/DTW approaches.
    • The findings are expected to generalize beyond 3D hand motion to other complex recognition tasks involving detailed measurements.