Applications of Convolution Neural Networks on Received Signal Strength Indicator Fingerprints for Indoor Localisation

This repo contains all the materials for my research project conducted under professor Gary Chan.

Abstract:

Indoor localisation is a hot topic in the field of big data and machine learning, which aims to utilise crowdsourced Received Signal Strength Indicator (RSSI) data collected from multiple reference access points (AP) propagating some variation of wave signal. This report addresses how attempts were made to utilise the power of well-studied convolutional neural networks (CNNs) in order to experiment with their methodologies and see if they could be applied to indoor localisation tasks using user location data in the form of these RSSI fingerprints. With the help of some state-of-the-art analytical tools, this report will demonstrate that there is potential for CNNs to play successful roles in indoor localisation tasks.

Introduction:

Indoor localisation refers to the task of locating an individual, or more specifically their devices on-hand, to gain some insight on their location in an indoor environment. Indoor localisation is an important task with a wide range of applications, such as intrusion detection for secure spaces , collecting commercial data on the shopping habits of customers inside a shopping mall, or even for monitoring the movement of infected individuals during the COVID-19 pandemic.