Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

Vitamin D alleviates intracerebral hemorrhage symptoms by modulating pro-inflammatory/anti-inflammatory phenotypic transformation of microglia through P2X7R activation of NLRP3.

IBRO neuroscience reports·2026
Same author

Brillouin microscopy for contact- and label-free quantification of tissue stiffness in lung fibrosis.

American journal of respiratory cell and molecular biology·2026
Same author

Comparative Genomics Analysis Reveals the Genomic Basis of S8 Proteases, CAZymes, and Secondary Metabolism Associated with Nematode Biocontrol in <i>Purpureocillium lilacinum</i>.

International journal of molecular sciences·2026
Same author

Clinical application of synovial fluid albumin-to-globulin ratio as a novel biomarker in diagnosing chronic periprosthetic joint infection.

Frontiers in cellular and infection microbiology·2026
Same author

The T2T genome assembly of Ziziphus jujuba 'Huizao' and pan-genome analyses provide insights into fruit texture diversity in jujube.

Molecular horticulture·2026
Same author

Evaluation of disinfection efficiency of new high oxygen membrane air disinfecting machine in hospital ward environment.

Scientific reports·2026

相关实验视频

Updated: May 10, 2025

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.4K

在使用DeepConv LSTM架构的边缘设备上高效地识别人类活动.

Haotian Zhou1, Xiujun Zhang2, Yu Feng1

  • 1School of Computer Science, Chengdu University, Chengdu, 610106, China.

Scientific reports
|April 22, 2025
PubMed
概括
此摘要是机器生成的。

在边缘设备上,TinyML可实现高效的人类活动识别 (HAR). 一个DeepConv LSTM模型在量子化后实现了97%的准确性,证明了TinyML.

关键词:
深度学习 (Deep Learning) 是一种深度学习.边缘计算 边缘计算人类活动识别 人类活动识别这就是为什么物联网是物联网物联网.量化模型的量化模型在TinyML中使用TinyML.

更多相关视频

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
06:49

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment

Published on: December 11, 2015

8.8K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

430

相关实验视频

Last Updated: May 10, 2025

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.4K
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
06:49

Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment

Published on: December 11, 2015

8.8K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

430

科学领域:

  • 人工智能的人工智能
  • 嵌入式系统 嵌入式系统
  • 机器学习 机器学习

背景情况:

  • 物联网 (IoT) 需要在资源有限的硬件上部署深度学习.
  • TinyML为边缘设备上的高效机器学习提供了一个解决方案.
  • 人类活动识别 (HAR) 对实时应用至关重要.

研究的目的:

  • 在边缘设备上使用TinyML部署轻量级的HAR深度学习模型.
  • 评估和比较不同深度学习架构对HAR的性能.
  • 为资源受限的部署优化一个模型.

主要方法:

  • 为HAR设计和评估了2D CNN,1D CNN和DeepConv LSTM模型.
  • 在表现最好的模型中应用完整整数量化.
  • 在使用Edge Impulse平台的Arduino Nano 33 BLE Sense Rev2上部署了量子化模型.

主要成果:

  • 在量子化之前,DeepConv LSTM实现了98.24%的准确性和98.23%的F1得分.
  • 量子化将模型大小从513.23KB减少到136.51KB.
  • 部署的模型保持了97%的准确性和97%的F1得分,内存 (29.1KB) 和闪存 (189.6KB) 使用量低.

结论:

  • 在边缘设备上,TinyML促进了高效和低延迟的HAR系统.
  • 在量子化后,DeepConv LSTM 模型适用于实时 HAR 应用.
  • 这项研究表明,在微控制器上部署先进的人工智能模型的实际可行性.