This repo contains the report (and perhaps the source code, if i ever get around to adding it) for the final project/dissertation I completed during the 3rd year of my computer science bachelor's at KCL.
The field of indoor localization is at a much different stage of development than seemingly similar, but much larger scale positioning systems. Such global positioning (GPS for example) has become an essential part of everyday life for many in the developed world, yet the thought of precise, accurate indoor positioning systems still seems a sci-fi-esque zenith of engineering. However, recent developments in both indoor-localization research itself, and the current boom in all things machine learning have lead to the emergence of some novel techniques that have brought the aforementioned zenith within our grasp.
This report deals with locating RFID tags, and will build upon an established technique of creating a fingerprint database from the signatures of known-location reference tags, by exploring the use of stacked denoising autoencoders, a class of deep neural networks, to perform the final comparison between the received signal of the desired tag and the reference tag fingerprints in the databas