Subject

Deep unfolding unrolls an optimization algorithm and maps the (sub)steps to corresponding neural network layers, to obtain a machine learning model that incorporates the domain knowledge from the original algorithm into its architecture. This approach results in very compact and efficient models, with many use cases in signal and image processing.

Integrated sensing and communication (ISAC) uses a wireless communication channel, for example WiFi or cellular networks, to also sense the environment through more elaborate processing of channel measurements. This dual use of the same hardware increases spectral and energy efficiency, and allows us to leverage existing communication infrastructure to help in e.g., robot localization or autonomous driving.

Kind of work

The student will explore the state-of-the-art in ISAC and get acquainted with the typical data types and processing steps. Next, the task is to evaluate and adapt the deep unfolding models from our research group to compressed sensing for ISAC, and apply them to micro-Doppler reconstruction and human activity recognition.

Framework of the Thesis

R. Mazzieri, J. Pegoraro, and M. Rossi, “Attention-Refined Unrolling for Sparse Sequential micro-Doppler Reconstruction,” IEEE Journal of Selected Topics in Signal Processing, vol. 18, 2024.

B. De Weerdt, Y. C. Eldar, and N. Deligiannis, “Deep Unfolding Transformers for Sparse Recovery of Video,” IEEE Transactions on Signal Processing, vol. 72, 2024.

Number of Students

1

Expected Student Profile

Knowledge of optimization problems, discrete Fourier transform

Excellent understanding of Python programming and machine learning

Motivation to seek a thorough understanding of deep unfolding and ISAC