Scholarlink

<p>Indoor human localization is crucial for multiple applications such as healthcare, energy management, and security. Compared with other device-free localization techniques like video, microwave radar and WiFi CSI, passive infrared (PIR) sensors offer advantages in privacy protection, low cost, low power consumption and electromagnetic interference resistance. However, most existing PIR-based methods relied on binary outputs of sensors and spatial partition, resulting in low localization accuracy. We developed and implemented a device-free system that integrates five analog PIR sensor modules to collect detailed signals for tracking individuals within a 4&#xa0;m <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\times\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>×</mo> </math></EquationSource> </InlineEquation> 4&#xa0;m area. We also create a high-quality dataset with many different walking paths, separate training and test sets, and accurate position labels. Using this data, we propose a method named MambaTrack-Net, which can accurately locate a person using just 0.1&#xa0;s of sensor data. Our experiments show that our method can find a person’s location with an average error of only 11&#xa0;cm, which is 52% better than the state-of-the-art method (Deep CNN-LSTM Network). Compared to current PIR indoor localization solutions, our approach introduces a clear technical advancement and addresses key challenges in low-cost, device-free, high-precision indoor positioning. The model’s real-time performance and suitability for parallel implementation highlight its potential for large-scale deployment in HPC-enabled smart 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

Mamba-empowered weak signals of passive infrared sensors for cost-effective indoor localization

  • 期刊论文

  • 发表日期：

    2026年2月25日

查看全文
  • Yan Kaiyu,
  • Pu Fangling,
  • Chen Hongjia,
  • Zhu Hengda,
  • Xu Xin

摘要

Indoor human localization is crucial for multiple applications such as healthcare, energy management, and security. Compared with other device-free localization techniques like video, microwave radar and WiFi CSI, passive infrared (PIR) sensors offer advantages in privacy protection, low cost, low power consumption and electromagnetic interference resistance. However, most existing PIR-based methods relied on binary outputs of sensors and spatial partition, resulting in low localization accuracy. We developed and implemented a device-free system that integrates five analog PIR sensor modules to collect detailed signals for tracking individuals within a 4 m \(\times\) × 4 m area. We also create a high-quality dataset with many different walking paths, separate training and test sets, and accurate position labels. Using this data, we propose a method named MambaTrack-Net, which can accurately locate a person using just 0.1 s of sensor data. Our experiments show that our method can find a person’s location with an average error of only 11 cm, which is 52% better than the state-of-the-art method (Deep CNN-LSTM Network). Compared to current PIR indoor localization solutions, our approach introduces a clear technical advancement and addresses key challenges in low-cost, device-free, high-precision indoor positioning. The model’s real-time performance and suitability for parallel implementation highlight its potential for large-scale deployment in HPC-enabled smart environments.