@inproceedings{29304, abstract = {{In this work we address disentanglement of style and content in speech signals. We propose a fully convolutional variational autoencoder employing two encoders: a content encoder and a style encoder. To foster disentanglement, we propose adversarial contrastive predictive coding. This new disentanglement method does neither need parallel data nor any supervision. We show that the proposed technique is capable of separating speaker and content traits into the two different representations and show competitive speaker-content disentanglement performance compared to other unsupervised approaches. We further demonstrate an increased robustness of the content representation against a train-test mismatch compared to spectral features, when used for phone recognition.}}, author = {{Ebbers, Janek and Kuhlmann, Michael and Cord-Landwehr, Tobias and Haeb-Umbach, Reinhold}}, booktitle = {{Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}}, pages = {{3860–3864}}, title = {{{Contrastive Predictive Coding Supported Factorized Variational Autoencoder for Unsupervised Learning of Disentangled Speech Representations}}}, year = {{2021}}, } @inproceedings{26770, abstract = {{Automatic transcription of meetings requires handling of overlapped speech, which calls for continuous speech separation (CSS) systems. The uPIT criterion was proposed for utterance-level separation with neural networks and introduces the constraint that the total number of speakers must not exceed the number of output channels. When processing meeting-like data in a segment-wise manner, i.e., by separating overlapping segments independently and stitching adjacent segments to continuous output streams, this constraint has to be fulfilled for any segment. In this contribution, we show that this constraint can be significantly relaxed. We propose a novel graph-based PIT criterion, which casts the assignment of utterances to output channels in a graph coloring problem. It only requires that the number of concurrently active speakers must not exceed the number of output channels. As a consequence, the system can process an arbitrary number of speakers and arbitrarily long segments and thus can handle more diverse scenarios. Further, the stitching algorithm for obtaining a consistent output order in neighboring segments is of less importance and can even be eliminated completely, not the least reducing the computational effort. Experiments on meeting-style WSJ data show improvements in recognition performance over using the uPIT criterion. }}, author = {{von Neumann, Thilo and Kinoshita, Keisuke and Boeddeker, Christoph and Delcroix, Marc and Haeb-Umbach, Reinhold}}, booktitle = {{Interspeech 2021}}, keywords = {{Continuous speech separation, automatic speech recognition, overlapped speech, permutation invariant training}}, title = {{{Graph-PIT: Generalized Permutation Invariant Training for Continuous Separation of Arbitrary Numbers of Speakers}}}, doi = {{10.21437/interspeech.2021-1177}}, year = {{2021}}, } @inproceedings{29173, author = {{von Neumann, Thilo and Boeddeker, Christoph and Kinoshita, Keisuke and Delcroix, Marc and Haeb-Umbach, Reinhold}}, booktitle = {{Speech Communication; 14th ITG Conference}}, location = {{Kiel}}, title = {{{Speeding Up Permutation Invariant Training for Source Separation}}}, year = {{2021}}, } @inproceedings{29308, abstract = {{In this paper we present our system for the Detection and Classification of Acoustic Scenes and Events (DCASE) 2021 Challenge Task 4: Sound Event Detection and Separation in Domestic Environments, where it scored the fourth rank. Our presented solution is an advancement of our system used in the previous edition of the task.We use a forward-backward convolutional recurrent neural network (FBCRNN) for tagging and pseudo labeling followed by tag-conditioned sound event detection (SED) models which are trained using strong pseudo labels provided by the FBCRNN. Our advancement over our earlier model is threefold. First, we introduce a strong label loss in the objective of the FBCRNN to take advantage of the strongly labeled synthetic data during training. Second, we perform multiple iterations of self-training for both the FBCRNN and tag-conditioned SED models. Third, while we used only tag-conditioned CNNs as our SED model in the previous edition we here explore sophisticated tag-conditioned SED model architectures, namely, bidirectional CRNNs and bidirectional convolutional transformer neural networks (CTNNs), and combine them. With metric and class specific tuning of median filter lengths for post-processing, our final SED model, consisting of 6 submodels (2 of each architecture), achieves on the public evaluation set poly-phonic sound event detection scores (PSDS) of 0.455 for scenario 1 and 0.684 for scenario as well as a collar-based F1-score of 0.596 outperforming the baselines and our model from the previous edition by far. Source code is publicly available at https://github.com/fgnt/pb_sed.}}, author = {{Ebbers, Janek and Haeb-Umbach, Reinhold}}, booktitle = {{Proceedings of the 6th Detection and Classification of Acoustic Scenes and Events 2021 Workshop (DCASE2021)}}, isbn = {{978-84-09-36072-7}}, pages = {{226–230}}, title = {{{Self-Trained Audio Tagging and Sound Event Detection in Domestic Environments}}}, year = {{2021}}, } @inproceedings{29306, abstract = {{Recently, there has been a rising interest in sound recognition via Acoustic Sensor Networks to support applications such as ambient assisted living or environmental habitat monitoring. With state-of-the-art sound recognition being dominated by deep-learning-based approaches, there is a high demand for labeled training data. Despite the availability of large-scale data sets such as Google's AudioSet, acquiring training data matching a certain application environment is still often a problem. In this paper we are concerned with human activity monitoring in a domestic environment using an ASN consisting of multiple nodes each providing multichannel signals. We propose a self-training based domain adaptation approach, which only requires unlabeled data from the target environment. Here, a sound recognition system trained on AudioSet, the teacher, generates pseudo labels for data from the target environment on which a student network is trained. The student can furthermore glean information about the spatial arrangement of sensors and sound sources to further improve classification performance. It is shown that the student significantly improves recognition performance over the pre-trained teacher without relying on labeled data from the environment the system is deployed in.}}, author = {{Ebbers, Janek and Keyser, Moritz Curt and Haeb-Umbach, Reinhold}}, booktitle = {{Proceedings of the 29th European Signal Processing Conference (EUSIPCO)}}, pages = {{1135–1139}}, title = {{{Adapting Sound Recognition to A New Environment Via Self-Training}}}, year = {{2021}}, } @inproceedings{35131, abstract = {{An FPGA accelerator for the computation of the semi-global Levenshtein distance between a pattern and a reference text is presented. The accelerator provides an important benefit to reduce the execution time of read-mappers used in short-read genomic sequencing. Previous attempts to solve the same problem in FPGA use the Myers algorithm following a column approach to compute the dynamic programming table. We use an approach based on diagonals that allows for some resource savings while maintaining a very high throughput of 1 alignment per clock cycle. The design is implemented in OpenCL and tested on two FPGA accelerators. The maximum performance obtained is 91.5 MPairs/s for 100 × 120 sequences and 47 MPairs/s for 300 × 360 sequences, the highest ever reported for this problem.}}, author = {{Castells-Rufas, David and Marco-Sola, Santiago and Aguado-Puig, Quim and Espinosa-Morales, Antonio and Moure, Juan Carlos and Alvarez, Lluc and Moreto, Miquel}}, booktitle = {{2021 31st International Conference on Field-Programmable Logic and Applications (FPL)}}, keywords = {{pc2-harp-ressources}}, publisher = {{IEEE}}, title = {{{OpenCL-based FPGA Accelerator for Semi-Global Approximate String Matching Using Diagonal Bit-Vectors}}}, doi = {{10.1109/fpl53798.2021.00036}}, year = {{2021}}, } @inproceedings{29937, author = {{Karp, Martin and Podobas, Artur and Jansson, Niclas and Kenter, Tobias and Plessl, Christian and Schlatter, Philipp and Markidis, Stefano}}, booktitle = {{2021 IEEE International Parallel and Distributed Processing Symposium (IPDPS)}}, publisher = {{IEEE}}, title = {{{High-Performance Spectral Element Methods on Field-Programmable Gate Arrays : Implementation, Evaluation, and Future Projection}}}, doi = {{10.1109/ipdps49936.2021.00116}}, year = {{2021}}, } @inproceedings{46194, author = {{Kenter, Tobias and Shambhu, Adesh and Faghih-Naini, Sara and Aizinger, Vadym}}, booktitle = {{Proceedings of the Platform for Advanced Scientific Computing Conference (PASC)}}, publisher = {{ACM}}, title = {{{Algorithm-hardware co-design of a discontinuous Galerkin shallow-water model for a dataflow architecture on FPGA}}}, doi = {{10.1145/3468267.3470617}}, year = {{2021}}, } @article{27099, abstract = {{In our work, we have engineered low capacitance single quantum dot photodiodes as sensor devices for the optoelectronic sampling of ultrafast electric signals. By the Stark effect, a time-dependent electric signal is converted into a time-dependent shift of the transition energy. This shift is measured accurately by resonant ps laser spectroscopy with photocurrent detection. In our experiments, we sample the laser synchronous output pulse of an ultrafast CMOS circuit with high resolution. With our quantum dot sensor device, we were able to sample transients below 20 ps with a voltage resolution in the mV-range.}}, author = {{Widhalm, Alex and Krehs, Sebastian and Siebert, Dustin and Sharma, Nand Lal and Langer, Timo and Jonas, Björn and Reuter, Dirk and Thiede, Andreas and Förstner, Jens and Zrenner, Artur}}, issn = {{0003-6951}}, journal = {{Applied Physics Letters}}, keywords = {{tet_topic_qd}}, pages = {{181109}}, title = {{{Optoelectronic sampling of ultrafast electric transients with single quantum dots}}}, doi = {{10.1063/5.0061358}}, volume = {{119}}, year = {{2021}}, } @unpublished{32236, abstract = {{The interaction between quantum light and matter is being intensively studied for systems that are enclosed in high-$Q$ cavities which strongly enhance the light-matter coupling. However, for many applications, cavities with lower $Q$-factors are preferred due to the increased spectral width of the cavity mode. Here, we investigate the interaction between quantum light and matter represented by a $\Lambda$-type three-level system in lossy cavities, assuming that cavity losses are the dominant loss mechanism. We demonstrate that cavity losses lead to non-trivial steady states of the electronic occupations that can be controlled by the loss rate and the initial statistics of the quantum fields. The mechanism of formation of such steady states can be understood on the basis of the equations of motion. Analytical expressions for steady states and their numerical simulations are presented and discussed.}}, author = {{Rose, H. and Tikhonova, O. V. and Meier, T. and Sharapova, P. }}, booktitle = {{arXiv:2109.00842}}, title = {{{Steady states of $Λ$-type three-level systems excited by quantum light in lossy cavities}}}, year = {{2021}}, } @unpublished{29655, author = {{Kirchgässner, Wilhelm and Wallscheid, Oliver and Böcker, Joachim}}, booktitle = {{arXiv preprint arXiv:2103.16323}}, title = {{{Thermal Neural Networks: Lumped-Parameter Thermal Modeling With State-Space Machine Learning}}}, year = {{2021}}, } @article{26283, author = {{Lüders, Carolin and Pukrop, Matthias and Rozas, Elena and Schneider, Christian and Höfling, Sven and Sperling, Jan and Schumacher, Stefan and Aßmann, Marc}}, issn = {{2691-3399}}, journal = {{PRX Quantum}}, title = {{{Quantifying Quantum Coherence in Polariton Condensates}}}, doi = {{10.1103/prxquantum.2.030320}}, year = {{2021}}, } @article{29748, author = {{Slawig, Diana and Gruschwitz, Markus and Gerstmann, Uwe and Rauls, Eva and Tegenkamp, Christoph}}, issn = {{1932-7447}}, journal = {{The Journal of Physical Chemistry C}}, keywords = {{Surfaces, Coatings and Films, Physical and Theoretical Chemistry, General Energy, Electronic, Optical and Magnetic Materials}}, number = {{36}}, pages = {{20087--20093}}, publisher = {{American Chemical Society (ACS)}}, title = {{{Adsorption and Reaction of PbPc on Hydrogenated Epitaxial Graphene}}}, doi = {{10.1021/acs.jpcc.1c06320}}, volume = {{125}}, year = {{2021}}, } @article{37331, abstract = {{High harmonic generation (HHG) from solids shows great application prospects in compact short-wavelength light sources and as a tool for imaging the dynamics in crystals with subnanometer spatial and attosecond temporal resolution. However, the underlying collision dynamics behind solid HHG is still intensively debated and no direct mapping relationship between the collision dynamics with band structure has been built. Here, we show that the electron and its associated hole can be elastically scattered by neighboring atoms when their wavelength approaches the atomic size. We reveal that the elastic scattering of electron/hole from neighboring atoms can dramatically influence the electron recombination with its left-behind hole, which turns out to be the fundamental reason for the anisotropic interband HHG observed recently in bulk crystals. Our findings link the electron/hole backward scattering with Van Hove singularities and forward scattering with critical lines in the band structure and thus build a clear mapping between the band structure and the harmonic spectrum. Our work provides a unifying picture for several seemingly unrelated experimental observations and theoretical predictions, including the anisotropic harmonic emission in MgO, the atomic-like recollision mechanism of solid HHG, and the delocalization of HHG in ZnO. This strongly improved understanding will pave the way for controlling the solid-state HHG and visualizing the structure-dependent electron dynamics in solids.}}, author = {{Zuo, Ruixin and Trautmann, Alexander and Wang, Guifang and Hannes, Wolf-Rüdiger and Yang, Shidong and Song, Xiaohong and Meier, Torsten and Ciappina, Marcelo and Duc, Huynh Thanh and Yang, Weifeng}}, issn = {{2765-8791}}, journal = {{Ultrafast Science}}, publisher = {{American Association for the Advancement of Science (AAAS)}}, title = {{{Neighboring Atom Collisions in Solid-State High Harmonic Generation}}}, doi = {{10.34133/2021/9861923}}, volume = {{2021}}, year = {{2021}}, } @article{23477, author = {{Thong, Le Huu and Ngo, Cong and Duc, Huynh Thanh and Song, Xiaohong and Meier, Torsten}}, issn = {{2469-9950}}, journal = {{Physical Review B}}, pages = {{085201}}, title = {{{Microscopic analysis of high harmonic generation in semiconductors with degenerate bands}}}, doi = {{10.1103/physrevb.103.085201}}, volume = {{103}}, year = {{2021}}, } @article{21946, abstract = {{Lithium niobate (LiNbO3), a material frequently used in optical applications, hosts different kinds of polarons that significantly affect many of its physical properties. In this study, a variety of electron polarons, namely free, bound, and bipolarons, are analyzed using first-principles calculations. We perform a full structural optimization based on density-functional theory for selected intrinsic defects with special attention to the role of symmetry-breaking distortions that lower the total energy. The cations hosting the various polarons relax to a different degree, with a larger relaxation corresponding to a larger gap between the defect level and the conduction-band edge. The projected density of states reveals that the polaron states are formerly empty Nb 4d states lowered into the band gap. Optical absorption spectra are derived within the independent-particle approximation, corrected by the GW approximation that yields a wider band gap and by including excitonic effects within the Bethe-Salpeter equation. Comparing the calculated spectra with the density of states, we find that the defect peak observed in the optical absorption stems from transitions between the defect level and a continuum of empty Nb 4d states. Signatures of polarons are further analyzed in the reflectivity and other experimentally measurable optical coefficients.}}, author = {{Schmidt, Falko and Kozub, Agnieszka L. and Gerstmann, Uwe and Schmidt, Wolf Gero and Schindlmayr, Arno}}, issn = {{2073-4352}}, journal = {{Crystals}}, pages = {{542}}, publisher = {{MDPI}}, title = {{{Electron polarons in lithium niobate: Charge localization, lattice deformation, and optical response}}}, doi = {{10.3390/cryst11050542}}, volume = {{11}}, year = {{2021}}, } @inproceedings{23474, author = {{Reichelt, Matthias and Rose, Hendrik and Kosarev, Alexander N. and Poltavtsev, Sergey V. and Bayer, Manfred and Akimov, Ilya A. and Schneider, Christian and Kamp, Martin and Höfling, Sven and Meier, Torsten}}, booktitle = {{Ultrafast Phenomena and Nanophotonics XXV}}, editor = {{Betz, Markus and Elezzabi, Abdulhakem Y.}}, title = {{{Controlling the emission time of photon echoes by optical freezing of exciton dephasing and rephasing in quantum-dot ensembles}}}, doi = {{10.1117/12.2576887}}, volume = {{11684}}, year = {{2021}}, } @article{22881, author = {{Nguyen, T. T. Nhung and Sollfrank, T. and Tegenkamp, C. and Rauls, E. and Gerstmann, Uwe}}, issn = {{2469-9950}}, journal = {{Physical Review B}}, pages = {{L201408}}, title = {{{Impact of screening and relaxation on weakly coupled two-dimensional heterostructures}}}, doi = {{10.1103/physrevb.103.l201408}}, volume = {{103}}, year = {{2021}}, } @article{22310, author = {{Neufeld, Sergej and Bocchini, Adriana and Schmidt, Wolf Gero}}, issn = {{2475-9953}}, journal = {{Physical Review Materials}}, title = {{{Potassium titanyl phosphate Z- and Y-cut surfaces from density-functional theory}}}, doi = {{10.1103/physrevmaterials.5.064407}}, year = {{2021}}, } @article{22008, author = {{Plaickner, Julian and Speiser, Eugen and Braun, Christian and Schmidt, Wolf Gero and Esser, Norbert and Sanna, Simone}}, issn = {{2469-9950}}, journal = {{Physical Review B}}, title = {{{Surface localized phonon modes at the Si(553)-Au nanowire system}}}, doi = {{10.1103/physrevb.103.115441}}, year = {{2021}}, }