Research on RSS Data Optimization and DFL Localization for Non-Empty Environments.

Area of Science:

• Signal Processing
• Machine Learning
• Wireless Communication

Background:

• Device-free localization (DFL) estimates location using radio frequency (RF) link shadowing.
• Complex environments introduce significant noise and multipath effects, degrading DFL accuracy.
• Existing filtering methods are often non-specific and yield unstable results.

Purpose of the Study:

• To enhance device-free localization (DFL) accuracy in complex environments.
• To address the limitations of conventional filtering techniques for received signal strength (RSS) data.
• To improve the generalization capabilities of machine learning models for DFL.

Main Methods:

• Proposed two-dimensional double correlation (TDDC) distributed wavelet filtering to denoise RSS data while preserving useful fluctuations.
• Developed an Adaboost.M2 ensemble learning model based on the Gini decision tree (GDTE) to enhance generalization for unknown environmental patterns.
• Conducted extensive experiments in two distinct drawing-room environments.

Main Results:

• TDDC filtering effectively removed random disturbances and noise from RSS data.
• The GDTE localization model demonstrated superior generalization ability compared to single models.
• Achieved high localization accuracy rates of 87% and 95% in the tested environments.

Conclusions:

• The combination of TDDC filtering and the GDTE model significantly improves DFL accuracy in complex settings.
• The proposed methods offer a robust solution for enhancing RSS data quality and localization performance.
• This research provides a valuable advancement for practical device-free localization applications.