Recent publications

3D-LaSR introduces a new way to reenact 3D human heads by navigating the latent space of state-of-the-art Gaussian-splatting GANs, enabling smooth, realistic animations and cross-identity expression transfer from just a single monocular video. By...

CGS-GAN enables, for the first time, stable, high-quality, and 3D-consistent head synthesis in 3D Gaussian Splatting GANs—without the view-conditioning that previously caused identity drift and inconsistencies. Through targeted multi-view...

This study presents a hyperspectral light field system for intraoperative tissue oxygen saturation (????????2) analysis and visualization. We present a correlation method for determining SO2 saturation with low computational demands. We...

Design and implementation of novel D-band (sub-THz) dual-channel (MIMO hybrid) beamforming wireless front-ends operating between 150 and 170 GHz. The system utilizes analog beamforming networks with two 16×16 microstrip Rotman lenses and dual...

Real-time demonstration of an ultra-broadband (4 GHz) integrated sensing and communications (ISAC) system operating at 160 GHz in the upper D-band. The demonstration showcased quasi-monostatic sensing and a 4.9 Gbit/s communication link,...

An enhanced digital self-interference cancellation technique for in-band full-duplex (FD) wireless communications, focusing on overcoming challenges from nonlinear distortions. The method uses a kernelized adaptive projected subgradient method...

An architecture for neuromorphic device-edge co-inference, with application to radio sensing for environment mapping in networked robotics scenarios. The developed demonstration setup integrates a radar sensor, radar preprocessing unit and...

This paper presents the development and implementation of a versatile ad-hoc metropolitan-range quantum key distribution (QKD) network. The approach presented integrates various types of physical channels and QKD protocols, and a mix of trusted...

The telecommunications industry is undergoing a paradigm shift toward open, disaggregated, and automated networks, necessitating a secure, regulated, and sovereign approach to telemetry data sharing among stakeholders. This paper introduces a...

Next generation coherent transceivers are expected to offer capacities of 1600 Gbps per wavelength, potentially requiring single-carrier symbol rates of 240 GBd or more. The analysis of higher order physical layer effects at these elevated symbol...

We have extended current 5G beamforming to reduce self-interference for integrated sensing and communication. Our principled beam codebook design follows novel analytical bounds to guarantee sensing performance alongside broadband...

Dielectric lens characterization in sub-THz frequencies requires not only high precision positioning of the feed antenna, but also the possibility to select multiple relative positions between the lens and the feed, especially in the case of...

We present a real-time stereo hyperspectral imaging system that combines 3D reconstruction with spectral tissue analysis, enabling live visualization of oxygenation and tissue composition during surgery.

We demonstrate modulation format identification on a micro ROADM ring using a photonic neural network with low-speed photodiodes.

In this paper, we introduce a novel time domain correlation based channel sounder that operates at 485 GHz.

This paper presents the design and implementation of an agile sub-terahertz (sub-THz) broadband time-domain photonic channel sounder, capable of characterizing wireless propagation channels within the 100 GHz to 500 GHz frequency range.

We introduce RIPE, an innovative reinforcement learning-based framework for weakly-supervised training of a keypoint extractor that excels in both detection and description tasks. In contrast to conventional training regimes that depend heavily...

This work introduces a two-stage waveform design framework for multi-antenna integrated sensing and communication systems. By optimizing radar beampatterns and communication signals under practical constraints, the proposed method achieves...

In modern 5G UEs, the PA consumes large amounts of power. However, there is a trade-off between power efficiency and nonlinear distortion. This study explores digital post-distortion to address PA nonlinearities at the base station. We take a...

The Open Optical Transport Network (OOPT) architecture, defined by the Telecom Infra Project (TIP), is currently being deployed by major operators. The OOPT ecosystem, with its multi-stakeholder and multi-vendor environment, faces challenges like...

Making sub-Terahertz frequencies usable in future mobile radio networks such as 6G requires adaptive beamforming techniques. Using newly developed D-band front ends, we present first experimental evidence for the feasibility of this approach.

This work introduces a computationally efficient method for MIMO radar beampattern synthesis through optimized waveform covariance matrix design. The approach directs transmitted power toward target regions, minimizes cross-correlations, and...

This paper investigates the potential to approximate a nonconvex beampattern design problem with constraints in the angular domain via a restriction to a convex subset of the feasible region. Simulations show that that imposing this convex...

In this paper, we present the Network Observability Platform (NOBS), a data-sovereign telemetry and monitoring framework designed to meet the primary needs of multi-vendor packet-optical networks in the Sixth Generation (6G) era. NOBS supports...

We experimentally demonstrate a photonic integrated circuit (PIC) that generates tunable mmWave signals with Hz-level linewidths. Utilizing on-chip optical frequency combs and injection locking, the phase-locked optical signals exhibit beat notes...

To enhance the resilience of network-cloud ecosystems, we establish a data governance framework for sharing optical testbed data across organizations and fostering machine learning research of optical networks. We further introduce multientity...

Active Learning is a powerful way to cut down the time and effort needed to annotate data, but its use in object detection remains underexplored. We introduce a Batch-Aware Active Learning (BAAL) framework that combines uncertainty with diversity...

Deploying deep learning models in clinical practice often requires leveraging multiple data modalities, such as images, text, and structured data, to achieve robust and trustworthy decisions. However, not all modalities are always available at...

Analysis of dielectric lens in sub-THz frequencies often demands huge computational effort for full-wave electromagnetic methods. This paper presents an asymptotic technique for lens analysis at D-band, using ray tracing and applying Kirchhoff’s...

This work investigates how smart placement of Reconfigurable Intelligent Surfaces (RISs) can extend wireless coverage by considering Line-of-Sight conditions and surface orientation as key design factors. A low-complexity algorithm enables...