Abstract:

In this talk, we utilize a variational autoencoder (VAE) for channel estimation (CE) in underdetermined (UD) systems. The basis of the approach forms a recently proposed concept in which a VAE is trained on channel state information (CSI) data. The VAE models the underlying unknown CSI data distribution as conditionally Gaussian, yielding the conditional first and second moments for the CSI, given a noisy observation. Afterward, the provided conditional moments are used to parameterize conditional linear minimum mean squared error (LMMSE) estimators that approximate the mean squared error (MSE)-optimal estimator by utilizing the VAE as a generative prior for the estimation problem. This work extends the existing framework from fully determined (FD) to UD systems, which are of high practical relevance. Particularly noteworthy is the extension of the estimator variant, which can be trained with noisy and compressed data and does not require perfect CSI during its offline training phase. This is a significant advantage compared to most other deep learning (DL)-based CE methods, where perfect CSI during the training phase is a crucial prerequisite. Numerical simulations demonstrate the superior performance of the proposed methods compared to related estimators. 

Biography:

Michael Baur is a fourth-year PhD student at the Technical University of Munich (TUM), Munich, Germany, and is currently a visiting researcher in IDCOM for September/October 2024. He received the B.Sc. and the M.Sc. (with high distinction) degrees in electrical engineering and information technology from TUM in 2017 and 2020, respectively. In 2020, he joined the Chair of Signal Processing with TUM, where he is currently a Research Assistant and PhD student. His research interests include generative models, statistical signal processing, and estimation theory, focussing on applications in wireless communications.