CCpos: WiFi Fingerprint Indoor Positioning System Based on CDAE-CNN

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• WiFi is a prevalent technology for indoor positioning due to its accessibility and range.
• Existing WiFi fingerprint localization methods face challenges with accuracy and robustness in complex indoor environments.
• Improving indoor localization accuracy is crucial for various applications, including navigation, asset tracking, and emergency services.

Purpose of the Study:

• To propose and evaluate a novel indoor WiFi positioning system, CCpos (CADE-CNN Positioning), designed to enhance accuracy and robustness.
• To leverage deep learning techniques, specifically a convolutional denoising autoencoder (CDAE) and a convolutional neural network (CNN), for improved WiFi-based localization.
• To assess the performance of CCpos against established datasets, demonstrating its effectiveness in real-world scenarios.

Main Methods:

• The CCpos system employs a two-stage approach: an offline stage for data preparation and an online stage for positioning.
• In the offline stage, the K-means algorithm is utilized to segment the training dataset for validation.
• In the online stage, a CDAE is used for Received Signal Strength Indicator (RSSI) denoising and feature extraction, followed by a CNN for location estimation.

Main Results:

• CCpos achieved mean positioning errors of 1.05 m on the Alcala Tutorial 2017 dataset and 12.4 m on the UJIIndoorLoc dataset.
• Experimental results validate the system's excellent noise immunity, indicating its reliability in environments with signal fluctuations.
• The proposed system demonstrates strong generalization performance, adapting effectively to different indoor environments and WiFi signal characteristics.

Conclusions:

• The CCpos system, integrating CDAE and CNN, significantly improves the accuracy and robustness of indoor WiFi fingerprint localization.
• The deep learning approach effectively handles noise and extracts relevant features from RSSI data, leading to superior positioning performance.
• CCpos offers a promising solution for accurate and reliable indoor positioning using widely available WiFi infrastructure.