Contributions by INUE at Asilomar Conference on Signals, Systems and Computers 2024

The Asilomar Conference on Signals, Systems, and Computers is an annual conference supported by the IEEE signal processing society held on the Asilomar conference grounds in Monterey, California. With around 400 participants, the conference covers a broad range of topics in signal processing, including machine learning, image processing, and communications. This year, the Institute of Telecommunications (INUE) was represented with three contributions on wireless communications:

• Uncertainty-Aware Dimensionality Reduction for Channel Charting with Geodesic Loss
• Hammerstein Self-Interference Cancellation for FDD in Adjacent Bands
• Towards Flexible LDPC Coding for 6G

The final papers will be available on IEEE Xplore around January 2025.

Channel Charting is a machine learning technique that aims to create a virtual map of the radio environment in a self-supervised manner. It exploits so-called Channel State Information (CSI), which is information about the wireless propagation channel and its multi-path properties. CSI is produced at the base station as a by-product of communicating with end users, so there is not need to change anything about the communication protocol and the system works with existing devices.

This year's presentation given at Asilomar was entitled "Uncertainty-Aware Dimensionality Reduction for Channel Charting with Geodesic Loss". It explains how to take the uncertainty about available information into account while training the channel chart. Furthermore, it proposes a solution to a fundamental issue that occurs when applying dimensionality reduction to datasets with non-convex low-dimensional representation, which is often an issue in the context of Channel Charting. With these suggested changes, localization performance could be significantly improved, especially under non-line of sight propagation conditions.

Ephraim Fuchs presented his work on "Hammerstein Self-Interference Cancellation for FDD in Adjacent Bands". In full-duplex systems that operate on neighboring frequency bands, leakage from the transmit band to the receive band introduces self-interference. To reduce reception errors due to self-interference, cancellation techniques can be deployed that use the current transmit signal to calculate and subtract a self-interference estimate. We present and compare three methods that follow a so-called Hammerstein model, a concatenation of a nonlinearity and a linear filter, to capture the out-of-band emissions from nonlinear components that propagate through a self-interference channel. Via simulation and measurements of an off-the-shelf WLAN device, we show that a proposed method outperforms the conventional method.

Historically, each new mobile standard has introduced major advances in channel coding and signal processing. These innovations have relied on new algorithms and implementations, but have often come at the expense of inter-generational compatibility. This year, in his presentation "Towards Flexible LDPC Coding for 6G", Paul Bezner explored a shift in code design towards a more unified approach using LDPC codes. The aim is to create a single code design that supports a wide range of parameters while minimising complexity in both specification and implementation. The paper proposes a new code design process that combines well-established methods, such as protograph-based raptor-like codes, with newer techniques, such as automorphism ensemble decoding. The result is a coding scheme that focuses on energy efficiency and reduced decoder complexity. It outperforms the 5G baseline while preserving essential, simple design principles.