Deep Learning-Based Device-Free Localization Scheme for Simultaneous Estimation of Indoor Location and Posture Using...

Area of Science:

• Internet-of-Things (IoT)
• Human Behavior Recognition
• Radar Systems

Background:

• Camera-based Device-Free Localization (DFL) is limited in dark environments and raises privacy concerns.
• Radar-based DFL offers a privacy-preserving alternative, effective in various lighting conditions.

Purpose of the Study:

• To propose a novel deep learning-based DFL scheme for simultaneous indoor location and posture estimation.
• To utilize 24-GHz frequency-modulated continuous-wave (FMCW) radars for enhanced DFL performance.

Main Methods:

• A parallel 1D convolutional neural network (CNN) architecture was employed.
• The CNN integrates regression for localization and classification for posture estimation (standing, sitting, lying, absence).
• Simultaneous 2D localization and posture recognition were achieved.

Main Results:

• The proposed scheme achieved an average localization error of 0.23 m, significantly outperforming conventional methods (0.65 m).
• Average posture estimation error was approximately 1.7%, vastly superior to conventional schemes (54.8% to 10%).

Conclusions:

• The deep learning-based FMCW radar DFL scheme enables accurate and simultaneous indoor localization and posture estimation.
• This approach overcomes limitations of camera-based systems, offering a robust, privacy-preserving solution for IoT applications.