Scholarlink

<p>The increasing demand for intelligent monitoring systems in elderly care has drawn significant attention to the development of non-invasive techniques for detecting falls and abnormal movements. This study presents a unified, privacy-preserving elderly monitoring framework designed for sensitive indoor environments (e.g., bathrooms) using Passive Infrared (PIR) sensors. The framework employs a sequential one-dimensional CNN-Bi-LSTM (1D-CNN-Bi-LSTM) architecture for coarse-grained occupancy-grid based human localization and fall detection. While existing PIR-based localization and tracking studies focus solely on position or trajectory estimation, the proposed framework extends PIR-based human localization toward localization-aware fall detection using bidirectional temporal modeling. The proposed approach avoids capturing identifiable visual information while effectively modeling temporal motion patterns from multi-sensor PIR data, making it particularly suitable for deployment in privacy-critical spaces where camera-based monitoring is not feasible. The proposed 1D-CNN-Bi-LSTM model was initially trained and cross-validated on a publicly available benchmark dataset for human localization, achieving 92% localization accuracy. Subsequently, a real-time dataset was curated using a five-sensor PIR setup to capture diverse indoor motion and fall scenarios, and the model was retrained on the custom real-time PIR dataset containing fall-event scenarios to perform localization-assisted fall detection, achieving 96% fall-detection accuracy. Extensive comparative evaluation with state-of-the-art sequential models demonstrates superior performance across multiple metrics like accuracy, precision, recall, and F1-Score. The results indicate that the proposed framework provides an effective and privacy-preserving solution for continuous elderly monitoring in constrained indoor environments.</p>

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

Deep learning-based localization and tracking of elderly individuals for healthcare monitoring

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月23日

查看全文
  • Priyanka Muruganandham,
  • Farjana Farvin Sahapudeen,
  • Sangeetha Jayaraman,
  • Ramesh Raman,
  • Vinay Kumar Reddy Chinthakunta Venkata Subba Reddy,
  • Purnendra Polisetti Nagaraju,
  • Surya Kiran Polisetti Nagaraju

摘要

The increasing demand for intelligent monitoring systems in elderly care has drawn significant attention to the development of non-invasive techniques for detecting falls and abnormal movements. This study presents a unified, privacy-preserving elderly monitoring framework designed for sensitive indoor environments (e.g., bathrooms) using Passive Infrared (PIR) sensors. The framework employs a sequential one-dimensional CNN-Bi-LSTM (1D-CNN-Bi-LSTM) architecture for coarse-grained occupancy-grid based human localization and fall detection. While existing PIR-based localization and tracking studies focus solely on position or trajectory estimation, the proposed framework extends PIR-based human localization toward localization-aware fall detection using bidirectional temporal modeling. The proposed approach avoids capturing identifiable visual information while effectively modeling temporal motion patterns from multi-sensor PIR data, making it particularly suitable for deployment in privacy-critical spaces where camera-based monitoring is not feasible. The proposed 1D-CNN-Bi-LSTM model was initially trained and cross-validated on a publicly available benchmark dataset for human localization, achieving 92% localization accuracy. Subsequently, a real-time dataset was curated using a five-sensor PIR setup to capture diverse indoor motion and fall scenarios, and the model was retrained on the custom real-time PIR dataset containing fall-event scenarios to perform localization-assisted fall detection, achieving 96% fall-detection accuracy. Extensive comparative evaluation with state-of-the-art sequential models demonstrates superior performance across multiple metrics like accuracy, precision, recall, and F1-Score. The results indicate that the proposed framework provides an effective and privacy-preserving solution for continuous elderly monitoring in constrained indoor environments.