Scholarlink

With the intensifying trend of population aging, the silver economy is flourishing. However, many elderly individuals face risks of secondary injuries during rehabilitation training due to the lack of professional guidance and affordable services. Consequently, developing a professional yet cost-effective rehabilitation detection system is urgently needed. This paper proposes MobileRehab, a smartphone-based rehabilitation training detection system that leverages the acoustic sensing capabilities of smartphones to generate acoustic motion images rich in motion information and robust to environmental noise. We constructed a dataset comprising over 9,000 samples and integrated a YOLOv12 object detection model enhanced with attention mechanisms and lightweight modules to achieve precise counting and classification of patients’ rehabilitation actions. Additionally, the system automatically evaluates quality metrics such as training rhythm and action stability, providing personalized rehabilitation guidance. Experimental results demonstrate that MobileRehab achieves an mAP50 of 96.9% across ten rehabilitation actions and exhibits excellent detection performance in various indoor environments, offering an efficient and convenient solution for home-based rehabilitation.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

MobileRehab: Smartphone-Based Rehabilitation Monitoring Using Active Acoustic Sensing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xiaolong Feng,
  • Kai Dong,
  • Xin Yang,
  • Hui Xia

摘要

With the intensifying trend of population aging, the silver economy is flourishing. However, many elderly individuals face risks of secondary injuries during rehabilitation training due to the lack of professional guidance and affordable services. Consequently, developing a professional yet cost-effective rehabilitation detection system is urgently needed. This paper proposes MobileRehab, a smartphone-based rehabilitation training detection system that leverages the acoustic sensing capabilities of smartphones to generate acoustic motion images rich in motion information and robust to environmental noise. We constructed a dataset comprising over 9,000 samples and integrated a YOLOv12 object detection model enhanced with attention mechanisms and lightweight modules to achieve precise counting and classification of patients’ rehabilitation actions. Additionally, the system automatically evaluates quality metrics such as training rhythm and action stability, providing personalized rehabilitation guidance. Experimental results demonstrate that MobileRehab achieves an mAP50 of 96.9% across ten rehabilitation actions and exhibits excellent detection performance in various indoor environments, offering an efficient and convenient solution for home-based rehabilitation.