Xue Wei
AboutThis is Wei Xue, I'm a Ph.D. candidate in the MESA lab, at University at Albany, SUNY. Between 2019 and 2020, I focused on robotics under the guidance of Dr. Weifu Wang. Since 2021, my research is on wireless communication, wireless steganography, RFI cancellation using RIS, deep learning, key generation, etc., supervised by Prof. Dola Saha. Before that, I received both MS and BS degrees from Xidian University in 2016 and 2019, respectively, and my main research interests are image processing and image segmentation. |
ResearchRIS circuit designAdvised by: Dr. Dola Saha and Dr. Aveek Dutta 2023 – present 
• Propose 3D RFI cancellation system by control the phase and amplitude to cancel incident RFI on telescope • Provide blueprints and circuit analysis for RIS array prototyping across multiple DoAs Multistage 2D DoA Estimation in Low SNRAdvised by: Dr. Dola Saha and Dr. Aveek Dutta 2022 – present 
• Propose a three stage algorithm that methodically exploits digital beamforming, creates virtual subarrays, inspects multiple options and introduces clustering to estimate the DoA in low SNRs RFI cancellation using RISAdvised by: Dr. Dola Saha and Dr. Aveek Dutta 2022 – present 
• Estimate incident RFI of the telescope and configure Reconfigurable Intelligent Surface(number, spacing, etc.) • Find the solution of amplitude and phase of each element of RIS to reflect incident RFI of RIS in order to cancel RFI in the telescope Key generationAdvised by: Dr. Dola Saha 2022 – present • Propose a Key Generation model using neural networks from wireless Channels • Extract the implicit features of channel in a compressed form to derive keys with high KGR and low KDR Wireless steganographyAdvised by: Dr. Dola Saha 2021 – present • Design a complex-valued neural network to hide the transmission of secret signals • This method has nothing to do with any properties of the covert signal (such as waveform or modulation order) • Use complex-valued neural networks, which is more powerful and has a faster learning speed than real-valued neural network • Transmit signals over the air and apply the transfer learning method to retrain the model again to further optimize the system and get a lower bit error rate Teaching human motionAdvised by: Dr. Weifu Wang 2019 – 2020
• Offer a systematic exploration of how to decompose complicated physical motions to make motion more interpretable and easier to learn • Build a teaching environment in a virtual environment (VR + unity) • Use real robot teaching frontend to demonstrate motion clips(sign language, breaststroke, butterfly, etc) |
PublicationsJournals:1. IDOL: Iterative Direction Of Arrival in Low SNR 1. WISE: Waveform Independent Signal Embedding for Covert Communication 2. MR Image Segmentation Based on Level Set Method Conferences:1. RIS for Signal Cancellation in 3D 2. KNEW: Key Generation using NEural Networks from Wireless Channels 3. Multistage 2D DOA Estimation in Low SNR 4. SCISRS: Signal Cancellation using Intelligent Surfaces for Radio Astronomy Services 5. KNEW: Key Generation using NEural Networks from Wireless Channels 6. Adversarial Learning for Hiding Wireless Signals 7. A Level Set Algorithm Based on Probabilistic Statistics for MR Image Segmentation 8. Collaborative Error Propagation for Single Sample Face Recognition |
AwardsACM SIGCOMM 2023 Student Travel Grant SIGCOMM2023 ACM WiSec2022 Student Travel Grants Wisec2022 Presidential Fellowship Award 2021 University at Albany Presidential Fellowship Award 2020 University at Albany Excellent Graduate Student Xidian University 2017 First Class Graduate Student Scholarship Xidian University 2016 Second Class Graduate Student Scholarship Xidian University 2013-2016 Undergraduate Scholarship Xidian University |