Research
I have worked as research assistant at LUMS from 2015 to 2018. During this period, I worked on RF sensing applications using channel state information (CSI) extracted from commercially available off-the-shelf WiFi radio (Intel 5300 WiFi Card).
I started experimentation with two laptops(Tx-Rx pair), each equipped with Intel 5300 WiFi card, for intrusion detection. I observed that whenever someone passes through the link between two laptops, while CSI is being captured at receiver, there are significant fluctuations in CSI data. It was my first hands-on experience with RF sensing. Later I worked on multiple applications relying on CSI and was able to publish two articles.
Here is a list of projects that I have worked on at LUMS;
- Preparation of sensing nodes using commercial off-the-shelf WiFi card (Intel 5300 WiFi Radio)
- Intrusion detection using channel state information (CSI) extracted from WiFi radio
- Through wall human motion detection by exploiting channel state information using horn antennas with WiFi radio
- Non-invasive monitoring and estimation of breathing rate with channel state information
- Non-obtrusive detection of concealed metallic object with WiFi radios
- Angle-of-arrival(AoA) and time-of-flight(ToF) estimation of WiFi signal for indoor localization
- Device-free indoor localization using joint AoA & ToF based fingerprinting
- Indoor localization using proximity sensing using Bluetooth low energy devices
In 2018, I joined ITU, Lahore for MS degree in electrical engineering. I was awarded graduate research fellowship. Having realized the potential of deep learning, I decided to pursue thesis that relies on concepts from machine learning and deep learning. During first year of MS, I worked on solving inverse problem (Fourier Ptychography) using deep learning model and have two papers that were accepted at ICCV 2019 and NeurIPS Workshop on Deep Learning and Inverse Problems, 2019. For my MS thesis, I worked on Bayesian network for root cause analysis (RCA) of faults in industrial plants and identification of most-influential-path between root-cause node and fault detection point.
As a research associate at ITU, Lahore, I have also worked on graph neural networks (GNNs) to learn expressive graph/node embeddings.