High Performance AE-CNN Model Enhanced via Gaussian Augmentation for RSSI-Dependent Indoor Positioning in Wireless Sensor Networks

Kahlaa K. Al-Nassrawy, Department of Information Network, College of Information Technology, University of Babylon, Babylon, IraqFollow
Ghaidaa A Al-Sultany, College of Engineering and IT, Al-Zahraa University for Women, Karbala, IraqFollow

Abstract

Precise indoor positioning system remains a significant challenge within Wireless Sensor Networks (WSNs) due to the instability and noise sensitivity of Received Signal Strength Indicator (RSSI) data. This paper proposes an advanced hybrid deep learning framework designed for 2D indoor localization, utilizing RSSI measurements to predict human or object position. Furthermore, to enhance the robustness of the model against signal variance and promote its generalization capability under this uncertainty, a data augmentation strategy is applied using three methods: Gaussian Noise Injection (GNI), Gaussian Mixture Model (GMM), and Bayesian Gaussian Mixture Model (BGMM). The purpose of data augmentation is to simulate actual noise and optimize the training dataset adaptability. High-accuracy architecture has been introduced that leverages an Auto-Encoder with a one-dimensional Convolutional Neural Network (AE–CNN). The Auto-Encoder acts as a noisy input filter and feature compressor, while the CNN acts as a spatial pattern extractor necessary for prediction. The original and augmented datasets were employed to train the proposed model. Its effectiveness was then tested utilizing error rate analysis and performance metrics. The best performance was achieved through Gaussian Noise Injection augmentation, yielding 98% accuracy and MSE corresponding to 0.48 m. This methodology offers a lightweight and noise-robust approach for location estimation, particularly in resource-constrained contexts where signal reliability is crucial.

Recommended Citation

Al-Nassrawy, Kahlaa K. and Al-Sultany, Ghaidaa A (2026) "High Performance AE-CNN Model Enhanced via Gaussian Augmentation for RSSI-Dependent Indoor Positioning in Wireless Sensor Networks," Karbala International Journal of Modern Science: Vol. 12 : Iss. 2 , Article 9.
Available at: https://doi.org/10.33640/2405-609X.3461