VariFi: Variable WiFi data rates during device-free sensing
Problem Statement:
Device-free sensing is the sensing of human activities or physiological variables (e.g. heart rate) using contactless solutions. These techniques achieve high accuracies in human activity recognition [1]. However, one important aspect is often neglected in current state-of-the-art: data transmission. These solutions often require the flooding of frequency bands with random packets, which is not a realistic scenario: often, data is only transmitted when it is available. This results in highly variable data rates, which are currently barely considered in device-free human activity recognition.
Task:
Your task is to use a state-of-the-art dataset collected using channel state information (radio wave propagation information) by the Pervasive Systems group and to analyse what happens when you variate the data rates (default Fs = 100 Hz). You will need to write a piece of code to simulate the variable data rates (MATLAB is preferred, but Python is fine, too). After this, the idea is that you look into how different sampling frequencies affect the accuracy (preferably deep learning), but also look into opportunities to deal with real-time classification with an unpredictable data stream. This assignment is great if you want to do data analysis with a state-of-the-art dataset collected by state-of-the-art technologies with an actual real-world application.
[1] Jeroen Klein Brinke and Nirvana Meratnia. 2019. Dataset: Channel state information for different activities, participants and days. In Proceedings of the 2nd Workshop on Data Acquisition To Analysis (DATA'19). Association for Computing Machinery, New York, NY, USA, 61–64. DOI:https://doi.org/10.1145/3359427.3361913
[2] https://github.com/seemoo-lab/nexmon_csi
Work:
10% implementation (software), 70% Data analysis, 20% Writing
Contact:
Jeroen Klein Brinke (j.kleinbrinke@utwente.nl)
Allow multiple students: YES