Convex Optimization via Symmetrical Hölder Divergence for a WLAN Indoor Positioning System

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• Indoor positioning systems (IPS) are crucial for services like emergency response and security.
• WiFi fingerprinting is a common IPS technique, but suffers from received signal strength (RSS) variance.
• Environmental changes cause RSS fluctuations, leading to multimodal distributions and inaccurate fingerprint databases.

Purpose of the Study:

• To propose a novel method for robust WiFi fingerprinting-based indoor localization.
• To address the challenge of RSS variance and multimodal signal distributions in indoor environments.
• To improve the accuracy and reliability of indoor positioning systems.

Main Methods:

• Developed a symmetrical Hölder divergence algorithm for WiFi signal analysis.
• Utilized both left-sided and right-sided data to symmetrize the centroid and minimize the algorithm.
• Applied statistical modeling of entropy, encompassing skew Bhattacharyya and Cauchy-Schwarz divergences.

Main Results:

• The proposed symmetrical Hölder divergence significantly outperformed traditional methods like k-nearest neighbor and probability neural networks.
• Achieved a position error accuracy of approximately 1 meter in experimental building environments.
• Demonstrated consistent performance improvements over existing indoor localization techniques.

Conclusions:

• Symmetrical Hölder divergence offers a robust solution for WiFi fingerprinting-based indoor localization.
• The method effectively mitigates the impact of RSS variance and environmental changes.
• This advancement holds potential for enhancing the precision and dependability of pervasive computing applications.