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

Attributes' Importance for Zero-Shot Pose-Classification Based on Wearable Sensors.

Hiroki Ohashi1, Mohammad Al-Naser2, Sheraz Ahmed3

  • 1Research & Development Group, Hitachi, Ltd., Tokyo 185-8601, Japan. hiroki.ohashi.uo@hitachi.com.

Sensors (Basel, Switzerland)
|August 4, 2018
PubMed
Summary

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

Unexpected intercostal fusion of total thoracic fracture dislocation with ankylosing spondylitis.

Surgical neurology international·2026
Same author

VariantMedium: sensitive and generalizable somatic point mutation calling with 3D DenseNets trained and evaluated on experimental data.

Genome medicine·2026
Same author

Current status of assisted peritoneal dialysis in Japan: A nationwide questionnaire survey.

Peritoneal dialysis international : journal of the International Society for Peritoneal Dialysis·2026
Same author

Explainable artificial intelligence in prostate cancer treatment recommendation: A decision support system for oncological expert panels.

European journal of cancer (Oxford, England : 1990)·2026
Same author

C-arm cone beam computed tomography with needle path overlay during fluoroscopic-guided screw placement for a spinal lesion.

BMC surgery·2026
Same author

FloraSyntropy-net: scalable deep learning with novel FloraSyntropy archive for large-scale plant disease diagnosis.

Plant methods·2026

This study introduces a novel zero-shot learning (ZSL) method that weights class attributes for improved accuracy. A new HDPoseDS dataset for wearable sensor-based pose classification is also presented, enhancing ZSL research.

Area of Science:

  • Machine Learning
  • Computer Vision
  • Wearable Technology

Background:

  • Zero-shot learning (ZSL) aims to classify unseen classes using auxiliary information like attributes.
  • Conventional ZSL methods often treat all attributes equally, leading to misclassification due to attribute relevance variance across classes.
  • This limitation can cause irrelevant attributes to negatively impact classification metrics for specific classes.

Purpose of the Study:

  • To propose a novel ZSL method that dynamically weights attribute importance for each class.
  • To introduce HDPoseDS, a comprehensive dataset for pose classification using wearable inertial measurement unit (IMU) sensors.
  • To evaluate the proposed ZSL method's performance using the new HDPoseDS dataset.

Main Methods:

  • A new ZSL approach is developed, calculating classification metrics by considering the importance of each attribute specific to each class.
Keywords:
CNNIMUaction recognitionpose classificationtime-serieswearable sensorzero-shot learning

Related Experiment Videos

  • A large-scale dataset, HDPoseDS, is created, featuring 22 pose classes from 10 subjects using 31 full-body IMU sensors.
  • The proposed ZSL method is benchmarked against existing methods on the HDPoseDS dataset.

    • Main Results:

    • The proposed ZSL method demonstrated superior performance compared to baseline methods.
    • A relative improvement of 5.9% was achieved over the best-performing baseline.
    • HDPoseDS is identified as a rich resource for zero-shot pose and action recognition research due to its sensor density.

    Conclusions:

    • The proposed attribute weighting method effectively enhances ZSL performance by addressing attribute relevance.
    • The HDPoseDS dataset provides a valuable resource for advancing research in wearable-based zero-shot recognition.
    • This work offers a significant step forward in both ZSL methodology and practical application in human pose analysis.