• Home
  • Search
  • Learning Sensor Sample-Reweighting for Dynamic Early-Exit Activity Recognition Via Meta Learning.
  • Cite Icon1
  • https://doi.org/10.1109/jbhi.2024.3524252Copy DOI Icon

Learning Sensor Sample-Reweighting for Dynamic Early-Exit Activity Recognition Via Meta Learning.

Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

During recent years, dynamic early-exit has provided a promising paradigm to improve the computational efficiency of deep neural networks by constructing multiple classifiers to let easy samples exit at shallow layers while avoiding redundant computations at deep exits, which has been seldom explored in the context of latency-aware human activity recognition (HAR) deployed on wearable devices. Particularly, most existing early-exit strategies have always treated all activity samples equally at each exit during training, which ignore such dynamic early-exit behavior at test-time, causing a potential mismatch between training and test. Intuitively, easy activity samples that often exit earlier at test-time should place more emphasis on the training loss of shallow classifiers, while hard activity samples should contribute more to the training loss of deep classifiers. To bridge this gap, this paper introduces a sample-reweighting approach for efficient activity inference, which employs a weight-predicting network to reweight the training loss of different activity samples at every exit. From a perspective of meta learning, a new optimization objective function is designed to jointly optimize both weight-predicting network and backbone network. We perform extensive experiments on three popular HAR benchmarks including UCI-HAR, WISDM, and UniMiB-SHAR, which demonstrate that while incorporating such test-time early-exit behavior into conventional training pipeline, it can consistently improve the accuracy-efficiency trade-offs under budgeted batch classification and anytime prediction patterns. Moreover, our approach has a natural advantage in handing class-imbalance HAR problem. Detailed ablation studies, visualized illustrations, and real hardware deployment are provided to support our statement.

Similar Papers
  • Conference Article
  • Citations9

Robust appearance-based human action recognition

  • Aug 23, 2004
  • Md Mizanur Rahman +1
  • Book Chapter
  • Citations1

A BRIEF INTRODUCTION TO HUMAN ACTIVITY RECOGNITION USING DEEP LEARNING

  • Dec 01, 2023
  • Tuhin Kumar Bera
  • Conference Article
  • Citations322

RGBD-HuDaAct: A color-depth video database for human daily activity recognition

  • Nov 01, 2011
  • Bingbing Ni +2
  • Conference Article
  • Citations136

Motion primitive-based human activity recognition using a bag-of-features approach

  • Jan 28, 2012
  • Mi Zhang +1
  • Research Article
  • Citations54

A Lightweight Framework for Human Activity Recognition on Wearable Devices

  • Nov 01, 2021
  • IEEE Sensors Journal
  • Yves Luduvico Coelho +3
  • Research Article
  • Citations1

XAttentionHAR Ensemble: Leveraging Cross-Modal Attention for Enhanced Activity Recognition

  • Jan 24, 2025
  • International Journal of Pattern Recognition and Artificial Intelligence
  • Sarita Sahni +2
  • PDF
  • Conference Article
  • Citations22

Mobeacon: An iBeacon-Assisted Smartphone-Based Real Time Activity Recognition Framework

  • Jan 01, 2015
  • Mohammad Arif Ul Alam +2
  • Research Article
  • Citations27

Fusing and refining convolutional neural network models for assembly action recognition in smart manufacturing

  • Jun 11, 2020
  • Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science
  • Md Al-Amin +6
  • PDF
  • Supplementary Content
  • Citations67

Channel State Information from Pure Communication to Sense and Track Human Motion: A Survey

  • Jul 29, 2019
  • Sensors (Basel, Switzerland)
  • Mohammed A A Al-Qaness +6
  • PDF
  • Research Article
  • Citations23

Device Position-Independent Human Activity Recognition with Wearable Sensors Using Deep Neural Networks

  • Mar 03, 2024
  • Applied Sciences
  • Sakorn Mekruksavanich +1
  • Research Article
  • Citations36

Spatio-Temporal Analysis for Human Action Detection and Recognition in Uncontrolled Environments

  • Jan 01, 2015
  • International Journal of Multimedia Data Engineering and Management
  • Dianting Liu +4
  • Conference Article
  • Citations6

An Investigation on Deep Learning-Based Activity Recognition Using IMUs and Stretch Sensors

  • Feb 21, 2022
  • Nguyen Thi Hoai Thu +1
  • Dissertation

Real-world applicable deep learning techniques

  • Jan 01, 2020
  • Ning Liu
  • Conference Article
  • Citations279

RadHAR

  • Oct 07, 2019
  • Akash Deep Singh +3
  • Research Article
  • Citations3

A Single Framework for Action Recognition Based on Boosted Randomized Trees

  • Jan 01, 2011
  • IPSJ Transactions on Computer Vision and Applications
  • Takayoshi Yamashita +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.