Adaptive Scheme of Denoising Autoencoder for Estimating Indoor Localization Based on RSSI Analytics in BLE...

Area of Science:

• Wireless communication
• Signal processing
• Indoor localization

Background:

• Indoor localization using Received Signal Strength Indicator (RSSI) is challenging due to signal noise from reflections and refractions.
• Noise in RSSI can be exponentially aggravated with distance, impacting localization accuracy.
• Existing methods struggle to effectively mitigate distance-dependent noise in RSSI signals.

Purpose of the Study:

• To enhance indoor localization performance by reducing noise in Bluetooth Low Energy (BLE) RSSI signals.
• To develop an adaptive noise generation scheme for training a denoising autoencoder (DAE) model.
• To improve the signal-to-noise ratio (SNR) as the distance between devices increases.

Main Methods:

• Utilized a denoising autoencoder (DAE) to process and denoise RSSI data from BLE signals.
• Proposed adaptive noise generation schemes to train the DAE, mimicking the exponential aggravation of noise with distance.
• Compared the DAE model's performance against Gaussian noise models and other localization algorithms, including the Kalman filter.

Main Results:

• Achieved a localization accuracy of 72.6%, representing a 10.2% improvement over a DAE model trained with Gaussian noise.
• Demonstrated superior denoising capabilities compared to the traditional Kalman filter.
• The adaptive noise generation effectively adapted to the characteristics of RSSI signal degradation.

Conclusions:

• The proposed adaptive noise generation scheme for DAE significantly improves indoor localization accuracy for BLE signals.
• This approach effectively addresses the challenge of distance-dependent noise in RSSI measurements.
• The DAE model with adaptive noise generation offers a promising solution for robust indoor positioning systems.