Recent publications

We present a novel approach that combines the high rendering quality of NeRF-based 3D-aware GANs with the flexibility and computational advantages of 3DGS. By training a decoder that maps implicit NeRF representations to explicit 3D Gaussian...

We analyze the system performance of a large bandwidth photonic THz transmitter based on a WR3 coupled photodiode. Employing probabilistic constellation shaping (PCS) alongside a 64-QAM base constellation, we achieve net data rates of up to 76.8...

We present a generative model that learns to synthesize human motion from limited training sequences. Our framework provides conditional generation and blending across multiple temporal resolutions. moOurdel adeptly captures human motion patterns...

How to provide robust connectivity against signal blockage in 6G mmWave/sub-THz networks? This study demonstrates that carefully designed multi-connectivity schemes can realize full macrodiversity gains and significant SNR gains through canonical...

We present a new automotive bus system which allows enhanced channel adaptation and fine-granular multi-user access based on OFDMA. We highlight advantages of OFDMA in automotive applications, sketch the concept and present an FPGA prototype.

We investigate the applicability of PAS for outdoor THz wireless links in simulations with weather-dependent loss models. Link performances are evaluated and optimal shaping entropies are determined to adjust error rates to a given FEC threshold....

We present here a public dataset for developing and testing ML models. The dataset is developed in a laboratory setting and includes 12672 samples including data points with different modulation formats, symbol rates, distances, WDM channel...

Nearly all diseases are caused by different combinations of exposures. We present the Causes of Outcome Learning approach (CoOL), which seeks to discover combinations of exposures that lead to an increased risk of a specific outcome in parts of...

We present SPVLoc, a global indoor localization method that accurately determines the six-dimensional (6D) camera pose of a query image and requires minimal scene-specific prior knowledge and no scene-specific training. Our approach employs a...

We introduce a compact scene representation organizing the parameters of 3D Gaussian Splatting (3DGS) into a 2D grid with local homogeneity, ensuring a drastic reduction in storage requirements without compromising visual quality during...

This report explores the future of grant-free random access in next-generation 3GPP wireless standards, focusing on how it aligns with the unsourced multiple access channel (UMAC) model.

Fraunhofer HHI researchers investigate the influence error of a fixed noise bandwidth assumption (i.e., 12.5 GHz) on optical signal-to-noise ratio (OSNR) measurements in the context of multi-band transmission and propose a correction method that...

We introduce BTSeg, an innovative, semi-supervised training approach enhancing semantic segmentation models in order to effectively handle a range of adverse conditions without requiring the creation of extensive new datasets. BTSeg employs a...

The D-band, spanning 110 GHz to 170 GHz, has emerged as a relevant frequency range for future mobile communications and radar sensing applications, particularly in the context of 6G technologies. This demonstration presents a high-bandwidth,...

In this paper, we introduce a novel time domain (TD) correlation based channel sounder that operates in the D-band (110 GHz to 170 GHz) with which we performed dual-polarized sub-THz channel measurements in an industrial environment.

Predicting user locations in dynamic indoor wireless environments is improved by using domain adaptation techniques. This approach effectively addresses challenges such as the birth, death, and drift of scattering clusters caused by environmental...

This paper presents the design and implementation of a novel analog beam former for the frequency range around 85 GHz. The capabilites are demonstrated through an outdoor trial in an urban environment, showcasing the potential of high frequency...

A method to interpret how restart-incremental bidirectional models process local ambiguities, with an analysis on meaning representation and dependency parsing.

We characterize the performance of high symbol rate probabilistic constellation shaping (PCS) in a THz wireless link utilizing a PIN photodiode-assisted transmitter. Using a 64-QAM base constellation we demonstrate high data rate adaptivity,...

The covariance matrix offers a balance between frequent channel estimations and maintaining acceptable system performance. In this work, we propose an intelligent autonomous mechanism that monitors changes in the covariance matrix, minimizing...

This paper reports the development outcomes of the world’s first integrated analog beamforming – transmit and receive – wireless front-ends operating in the D-band (a sub- THz band), particularly designed to operate within 150 GHz to 170 GHz...

Discover a groundbreaking two-timescale method for optimizing the weighted sum-rate in future multi-antenna wireless systems. Our innovative approach overcomes the severe scalability issues of traditional methods, namely the extreme computational...

We propose that deploying Reflecting Intelligent Surfaces (RIS) as auxiliary sensors can either enable (in cases of blockage) or assist UEs in improving sensing performance. This can be achieved by deploying larger surfaces, offering superior...

Efficiently managing network load has long been a challenge in wireless networks, and open radio access networks (O-RAN) can offer a potential solution through open interfaces and optimization capabilities. This paper focuses on addressing the...

We investigated the transmission properties of optical waveguide based on fluorinated acrylate polymer and Ormocer® based polymer in the visible range (VIS). Laser transmission-induced transparency (LTIT) and fluorescence were observed in the...

Neurons in deep neural networks can act polysemantically, meaning that they encode for multiple (unrelated) features. As such, understanding the inner workings of machine learning models becomes more difficult. We present PURE to turn...

While recent studies show high accuracy in the classification of Alzheimer's disease using deep neural networks, the underlying learned concepts have not been investigated. We separated Alzheimer's patients (n=117) from normal controls (n=219) by...

DNNs are prone to relying on spurious correlations in data, posing risks in critical applications. Post-hoc methods exist to mitigate this without retraining but can globally shift latent features distributions, harming model performance. We...

With PCX, we introduce a method that summarizes similar single explanations via prototypical ones. As such, we can understand the whole model behavior quickly and in detail. PCX further allows to validate individual predictions by communicating...

Our new method, AttnLRP, is the first to faithfully and holistically attribute not only input but also latent representations of transformer models with the computational efficiency similar to a single backward pass. We demonstrate that our...