WiFi Fingerprinting Indoor Localization Based on Dynamic Mode Decomposition Feature Selection with Hidden Markov...

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• WiFi Received Signal Strength Indicator (RSSI) measurements are crucial for indoor localization where GPS is unavailable.
• RSSI measurements are nonlinear and susceptible to indoor environmental interference, necessitating advanced processing.
• Traditional machine learning methods like Hidden Markov Models (HMM) for WiFi fingerprinting face challenges with high-dimensional data and computational costs.

Purpose of the Study:

• To introduce a novel feature extraction method using Dynamic Mode Decomposition (DMD) to enhance Hidden Markov Model (HMM) based WiFi indoor localization.
• To reduce the feature dimension of RSSI data for more efficient HMM processing.
• To improve the accuracy and reduce the computational cost of WiFi fingerprinting localization.

Main Methods:

• Feature extraction using Dynamic Mode Decomposition (DMD) to decompose RSSI measurements into meaningful spatial and temporal components.
• Analytical reconstruction of DMD mode forms to generate low-dimensional feature vectors.
• Integration of these low-dimensional features into a Hidden Markov Model (HMM) for indoor localization.

Main Results:

• The proposed HMM-DMD algorithm demonstrated significant improvements in localization performance.
• The method achieved higher accuracy compared to existing state-of-the-art machine learning algorithms for WiFi fingerprinting.
• The HMM-DMD approach offers a reasonable processing time, balancing performance and computational efficiency.

Conclusions:

• Dynamic Mode Decomposition (DMD) effectively extracts meaningful features from WiFi RSSI data for indoor localization.
• The HMM-DMD combination provides a superior approach to WiFi fingerprinting, enhancing accuracy and efficiency.
• This method represents a significant advancement in indoor positioning systems utilizing WiFi signals.