Multi-Dimensional Wi-Fi Received Signal Strength Indicator Data Augmentation Based on Multi-Output Gaussian Process...

Area of Science:

• Computer Science
• Artificial Intelligence
• Signal Processing

Background:

• Indoor localization commonly uses Received Signal Strength Indicators (RSSIs) with Wi-Fi infrastructure.
• Deep Neural Networks (DNNs) enhance indoor localization accuracy but require extensive labeled training data.
• Data collection for DNNs is time-consuming and challenging, especially in large-scale environments or during pandemics.

Purpose of the Study:

• To address data scarcity in indoor localization by investigating multi-dimensional RSSI data augmentation.
• To leverage Multi-Output Gaussian Process (MOGP) for exploiting correlations among RSSIs from multiple access points.
• To enhance the performance of state-of-the-art indoor localization models using augmented data.

Main Methods:

• Investigated multi-dimensional RSSI data augmentation using Multi-Output Gaussian Process (MOGP).
• MOGP was employed to exploit correlations among RSSIs across multiple access points, floors, and buildings.
• Experiments utilized a hierarchical Recurrent Neural Network (RNN) model and the UJIIndoorLoc database.

Main Results:

• MOGP-based data augmentation demonstrated feasibility for improving indoor localization.
• The RNN model trained with MOGP-augmented data, specifically using the 'by a single building' mode, outperformed other augmentation methods.
• This approach achieved a three-dimensional localization error of 8.42 meters.

Conclusions:

• MOGP-based RSSI data augmentation effectively addresses data collection challenges for indoor localization.
• Exploiting correlations among RSSIs via MOGP enhances the accuracy of Wi-Fi fingerprinting.
• The proposed method offers a viable solution for large-scale, reliable indoor localization systems.