HKU FYP
Deep Learning-Powered Wi-Fi Fingerprinting for Scalable Indoor Localization

Han Xiaoyuan

Introduction

In general, the Wi-Fi fingerprint indoor positioning process usually includes two phases: offline and online. In the offline phase, the fingerprint database is designed to measure and record Wi-Fi signals from different access points (APs) with the received signal strength (RSS). During the online phase, the measurement of the user’s RSS is calculated using classificational algorithms with the fingerprints in the database.

Fig 1. Overview of Wi-Fi Fingerprinting

Progresses

Tentative input for the deep learning model include:

Reference point locations
Wi-Fi access point locations
Received Signal Strength Indicator
Floor map

RSSI collected

Collected at main Library 1st Floor
4590 data samples
More data collection in progress…

Fig 2. RSSI data overview. The horizontal axis represent the index of the APs (d = 17), and the vertical axis represents the index of the reference points (c = 27).

Fig 3. Floor plan draft

Floor plan

Preprocessing required
Red dots – APs
Blue dots – RPs

A passion for creating spaces

Our comprehensive suite of professional services caters to a diverse clientele, ranging from homeowners to commercial developers.