TY - CHAP A1 - Łopatka, Kuba A1 - Bocklet, Tobias T1 - State Sequence Pooling Training of Acoustic Models for Keyword Spotting BT - Proceedings of the Annual Conference of the International Speech Communication Association T2 - Interspeech 2020 N2 - We propose a new training method to improve HMM-based keyword spotting. The loss function is based on a score computed with the keyword/filler model from the entire input sequence. It is equivalent to max/attention pooling but is based on prior acoustic knowledge. We also employ a multi-task learning setup by predicting both LVCSR and keyword posteriors. We compare our model to a baseline trained on frame-wise cross entropy, with and without per-class weighting. We employ a low-footprint TDNN for acoustic modeling. The proposed training yields significant and consistent improvement over the baseline in adverse noise conditions. The FRR on cafeteria noise is reduced from 13.07% to 5.28% at 9 dB SNR and from 37.44% to 6.78% at 5 dB SNR. We obtain these results with only 600 unique training keyword samples. The training method is independent of the frontend and acoustic model topology. KW - keyword spotting KW - machine learning KW - speech recognition Y1 - 2020 U6 - https://doi.org/https://doi.org/10.21437/interspeech.2020-2722 SN - 2958-1796 SP - 4338 EP - 4342 PB - ISCA CY - ISCA ER - TY - CHAP A1 - Georges, Munir A1 - Huang, Jonathan A1 - Bocklet, Tobias T1 - Compact Speaker Embedding: lrx-Vector BT - Proceedings of the Annual Conference of the International Speech Communication Association T2 - Interspeech 2020 N2 - Deep neural networks (DNN) have recently been widely used in speaker recognition systems, achieving state-of-the-art performance on various benchmarks. The x-vector architecture is especially popular in this research community, due to its excellent performance and manageable computational complexity. In this paper, we present the lrx-vector system, which is the low-rank factorized version of the x-vector embedding network. The primary objective of this topology is to further reduce the memory requirement of the speaker recognition system. We discuss the deployment of knowledge distillation for training the lrx-vector system and compare against low-rank factorization with SVD. On the VOiCES 2019 far-field corpus we were able to reduce the weights by 28% compared to the full-rank x-vector system while keeping the recognition rate constant (1.83% EER). KW - speaker recognition KW - x-vector KW - low power Y1 - 2020 U6 - https://doi.org/https://doi.org/10.21437/interspeech.2020-2106 SN - 2958-1796 SP - 3236 EP - 3240 PB - ISCA CY - ISCA ER - TY - CHAP A1 - Chen, Wenda A1 - Huang, Jonathan A1 - Bocklet, Tobias T1 - Length- and Noise-Aware Training Techniques for Short-Utterance Speaker Recognition T2 - Interspeech 2020 N2 - Speaker recognition performance has been greatly improved with the emergence of deep learning. Deep neural networks show the capacity to effectively deal with impacts of noise and reverberation, making them attractive to far-field speaker recognition systems. The x-vector framework is a popular choice for generating speaker embeddings in recent literature due to its robust training mechanism and excellent performance in various test sets. In this paper, we start with early work on including invariant representation learning (IRL) to the loss function and modify the approach with centroid alignment (CA) and length variability cost (LVC) techniques to further improve robustness in noisy, far-field applications. This work mainly focuses on improvements for short-duration test utterances (1-8s). We also present improved results on long-duration tasks. In addition, this work discusses a novel self-attention mechanism. On the VOiCES far-field corpus, the combination of the proposed techniques achieves relative improvements of 7.0% for extremely short and 8.2% for full-duration test utterances on equal error rate (EER) over our baseline system. KW - speaker recognition KW - invariant representation learning KW - centroid alignment KW - x-vector KW - far-field Y1 - 2020 U6 - https://doi.org/https://doi.org/10.21437/interspeech.2020-2872 SN - 2958-1796 SP - 3835 EP - 3839 PB - ISCA CY - ISCA ER - TY - CHAP A1 - Vasquez, Juan A1 - Bocklet, Tobias A1 - Orozco, JuanRafael A1 - Nöth, Elmar T1 - Comparison of user models based on GMM-UBM and i-vectors for speech, handwriting, and gait assessment of Parkinson’s disease patients N2 - Parkinson's disease is a neurodegenerative disorder characterized by the presence of different motor impairments. Information from speech, handwriting, and gait signals have been considered to evaluate the neurological state of the patients. On the other hand, user models based on Gaussian mixture models - universal background models (GMM-UBM) and i-vectors are considered the state-of-the-art in biometric applications like speaker verification because they are able to model specific speaker traits. This study introduces the use of GMM-UBM and i-vectors to evaluate the neurological state of Parkinson's patients using information from speech, handwriting, and gait. The results show the importance of different feature sets from each type of signal in the assessment of the neurological state of the patients. KW - Parkinson’s disease, GMM-UBM, i- vectors, gait analysis, handwriting analysis, speech analysis Y1 - 2020 U6 - https://doi.org/10.48550/arXiv.2002.05412 SP - 6544 EP - 6548 ER - TY - JOUR A1 - Schuster, Maria A1 - Arias-Vergara, Tomas A1 - Müller-Hörner, Rainer A1 - Winterholler, Cordula A1 - Bocklet, Tobias T1 - Verstehen mich mit der Maske noch alle? T1 - Speech quality changes due to face masks BT - Coronavirus-Pandemie JF - MMW - Fortschritte der Medizin N2 - In Arztpraxen nimmt der Infektionsschutz mittels Mund-Nasen-Maske momentan eine zentrale Rolle ein. Wenig nachgedacht wird dabei über die Folgen für die Kommunikation. Gerade schwerhörigen Patienten bereiten gedämpfte Stimmen und abgedeckte Lippen Probleme. KW - Face mask, speech quality, computer-assisted speech assessment, hearing loss Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1007/s15006-020-0749-4 VL - 162 IS - 14 SP - 42 EP - 44 ER - TY - CHAP A1 - Lopatka, Kuba A1 - Bocklet, Tobias T1 - State Sequence Pooling Training of Acoustic Models for Keyword Spotting T2 - Proceedings Interspeech 2020 N2 - We propose a new training method to improve HMM-based keyword spotting. The loss function is based on a score computed with the keyword/filler model from the entire input sequence. It is equivalent to max/attention pooling but is based on prior acoustic knowledge. We also employ a multi-task learning setup by predicting both LVCSR and keyword posteriors. We compare our model to a baseline trained on frame-wise cross entropy, with and without per-class weighting. We employ a low-footprint TDNN for acoustic modeling. The proposed training yields significant and consistent improvement over the baseline in adverse noise conditions. The FRR on cafeteria noise is reduced from 13.07% to 5.28% at 9 dB SNR and from 37.44% to 6.78% at 5 dB SNR. We obtain these results with only 600 unique training keyword samples. The training method is independent of the frontend and acoustic model topology. KW - keyword spotting, machine learning, speech recognition Y1 - 2020 U6 - https://doi.org/10.21437/Interspeech.2020-2722 SN - 2958-1796 SP - 4338 EP - 4342 ER - TY - CHAP A1 - Georges, Munir A1 - Huang, Jonathan A1 - Bocklet, Tobias T1 - Compact Speaker Embedding BT - lrx-vector N2 - Deep neural networks (DNN) have recently been widely used in speaker recognition systems, achieving state-of-the-art performance on various benchmarks. The x-vector architecture is especially popular in this research community, due to its excellent performance and manageable computational complexity. In this paper, we present the lrx-vector system, which is the low-rank factorized version of the x-vector embedding network. The primary objective of this topology is to further reduce the memory requirement of the speaker recognition system. We discuss the deployment of knowledge distillation for training the lrx-vector system and compare against low-rank factorization with SVD. On the VOiCES 2019 far-field corpus we were able to reduce the weights by 28% compared to the full-rank x-vector system while keeping the recognition rate constant (1.83% EER). KW - speaker recognition, x-vector, low power Y1 - 2020 U6 - https://doi.org/10.48550/arXiv.2008.05011 ER - TY - JOUR A1 - Schkutow, Andreas A1 - Scholle, Karsten A1 - Fuhrberg, Peter A1 - Frick, Thomas T1 - Scanning techniques for optimized damage tolerance in quasi-simultaneous laser transmission welding of plastics JF - Procedia CIRP N2 - Fiber lasers are increasingly replacing the commonly used diode lasers in laser transmission welding of plastics for challenging applications due to their superior beam quality and the ability to use long working distances and small spot sizes. At the same time, these smaller spot sizes increase the risk of thermal degradation. In this work, we investigate different scanning strategies that allow controlling the weld seam widths. These strategies are characterized regarding the demands on the performance of the scanner system, the risk of thermal degradation and the mechanical properties of the resulting welds. KW - aser transmission welding; quasi-simultaneous welding; irradiation strategies; offset-scanning technique; thermal degradation Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1016/j.procir.2020.09.120 SN - 2212-8271 VL - 94 SP - 697 EP - 701 PB - Elsevier BV ER - TY - JOUR A1 - Ebinger, Frank A1 - Omondi, Bramwel T1 - Leveraging Digital Approaches for Transparency in Sustainable Supply Chains BT - A Conceptual Paper JF - Sustainability N2 - With the growing global interdependence of companies, their scope of responsibility for the environmental, social, and human rights impacts associated with their activities is also growing. In this context, companies are increasingly held accountable for social and ecological issues that lie within the sphere of action of their suppliers and sub-suppliers. They are thus faced with the challenge of meeting these demands for transparency, traceability, and compliance with standards in their Supply Chains (SCs). Based on the theoretical framework of Sustainable Supply Chain Transparency (SSCT) in Sustainable Supply Chain Management (SSCM), this conceptual article aims at initiating the discussion on digitalization in SSCM. Therefore, a heuristical research framework, based on relevant fields in the management of sustainability oriented transparency (governance, cooperation and partner selection, traceability/tracking, and strategic and operational risk assessment) is developed. In relation to these fields, consequently, data-driven digital approaches are identified to which potentials for SSCT and control can be attributed. This initial analysis of existing digital approaches already shows that the market is developing dynamically, but is driven more by individual initiatives. In many cases, the approaches used so far are still in the trial phase or offer only limited solutions. Therefore, this paper contributes by giving an overview of the current application of the digitalization approaches in SSCM pinning our discussion on SSCT dimensions. KW - sustainable supply chain management; data-driven technologies; sustainable supply chain transparency; blockchain technology; artificial intelligence; cloud-computing Y1 - 2020 U6 - https://doi.org/https://doi.org/10.3390/su12156129 SN - 2071-1050 VL - 12 IS - 15 PB - MDPI AG ER - TY - JOUR A1 - Dametew, Alie Wube A1 - Beshah, Birehanu A1 - Ebinger, Frank ED - Zhou, Zude T1 - The challenges and practice of metal industries into global supply chain integration BT - A literature review JF - Cogent Engineering N2 - The objective of this study was to investigate the current research practices and existing progress research published on supply chain management and supply chain integration in an attempt to provide a research agenda for future investigations. Mainly the secondary data analysis and methods employed for the investigations. Particularly a systematic literature review analysis from 115 papers published from 16 important journals in the supply chain, logistics, Industrial management & data systems, supply chain management: an international journal and operations management during the period 1997 to 2018 was employed. This paper has identified relevant supply chain integration research studies that have contributed to the development and growth of manufacturing industries performance and competitiveness to the SCI and SCM fields. The analysis of the relevant articles gaps and miss-concepts in the previous literature provides to reduce universality and insufficient ordinary view, confusion of supply chain integrations were investigated. The gaps and investigations were done based on their methodological porches, objectives, analysis and research finding. Mainly the gaps, miss concepts and challenges concepts in the previous literature were identified and grouped in five issues. Moreover, the gaps and recommendations for future research are consolidated into a research agenda that provides practitioners to evaluate existing SCI & SCM issues and seek ways to develop a conceptual farm work on SCI of manufacturing industries to global supply chains. Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1080/23311916.2020.1762523 SN - 2331-1916 VL - 7 IS - 1 PB - Informa UK Limited ER - TY - JOUR A1 - Fuhrberg, Peter A1 - Ahrens, Anja A1 - Schkutow, Andreas A1 - Frick, Thomas T1 - 2.0 μm Laser Transmission Welding BT - Welding of transparent and opaque polymers with single‐mode Tm‐doped fiber lasers JF - PhotonicsViews N2 - 2.0 μm fiber lasers provide a high beam quality and a high‐power output, which makes them ideal for welding and cutting a wide variety of commercially used plastics, as well as marking plastics, metals or even food. At 2.0 μm wavelength the intrinsic absorption of most thermoplastics is high enough to weld or cut without applying any absorbent additives or coatings. Compared to commonly used near‐infrared systems, welding with 2.0 μm lasers can improve heat distribution and gap bridging and enable transmission welding for challenging materials. Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1002/phvs.202000013 SN - 2626-1294 VL - 17 IS - 2 SP - 64 EP - 68 PB - Wiley ER - TY - CHAP A1 - Brost, Julian A1 - Egger, Christoph A1 - Lai, Russell A1 - Schmid, Fritz A1 - Schröder, Dominique A1 - Zoppelt, Markus T1 - Threshold Password-Hardened Encryption Services N2 - Password-hardened encryption (PHE) was introduced by Lai et al. at USENIX 2018 and immediately productized by VirgilSecurity. PHE is a password-based key derivation protocol that involves an oblivious external crypto service for key derivation. The security of PHE protects against offline brute-force attacks, even when the attacker is given the entire database. Furthermore, the crypto service neither learns the derived key nor the password. PHE supports key-rotation meaning that both the server and crypto service can update their keys without involving the user. While PHE significantly strengthens data security, it introduces a single point of failure because key-derivation always requires access to the crypto service. In this work, we address this issue and simultaneously increase security by introducing threshold password-hardened encryption. Our formalization of this primitive revealed shortcomings of the original PHE definition that we also address in this work. Following the spirit of prior works, we give a simple and efficient construction using lightweight tools only. We also implement our construction and evaluate its efficiency. Our experiments confirm the practical efficiency of our scheme and show that it is more efficient than common memory-hard functions, such as scrypt. From a practical perspective this means that threshold PHE can be used as an alternative to scrypt for password protection and key-derivation, offering better security in terms of offline brute force attacks. CCS CONCEPTS • Security and privacy → Cryptography. Y1 - 2020 U6 - https://doi.org/10.1145/3372297.3417266 SP - 409 EP - 424 PB - ACM ER - TY - CHAP A1 - Wendel, Sebastian A1 - Karamanakos, Petros A1 - Dietz, Armin A1 - Kennel, Ralph T1 - Flux Linkage-Based Model Predictive Current Control for Nonlinear PMSM Drives T2 - IECON 2020 The 46th Annual Conference of the IEEE Industrial Electronics Society N2 - In this paper, a flux linkage-based direct model predictive current control approach is presented for small permanent magnet synchronous motor (PMSM) drives. The method aims to minimize the current ripples at steady state by deciding on the optimal switching instant, while exhibiting fast dynamic behavior during transients. To this end, the future trajectory of the stator current is not computed based on the machine inductances or inductance look-up tables, but on the changes of the magnetic flux linkage by utilizing flux linkage maps. As shown, the proposed method can be particularly advantageous for electric drives with a noticeable nonlinearity in terms of saturation and/or cross-coupling effects since it allows for a significantly increased prediction accuracy, which leads to an improved steady-state performance as indicated by the reduced current distortions. KW - Couplings ; Switches; Predictive models ; Mathematical model ; Rotors ; Real-time systems ; Prediction algorithmso Y1 - 2022 U6 - https://doi.org/https://doi.org/10.1109/iecon43393.2020.9254484 PB - IEEE ER - TY - CHAP A1 - Gerstner, Michael A1 - Maerz, Martin A1 - Dietz, Armin T1 - Reconfigurable Transformer for Wide Input Voltage Range DC/DC Converters T2 - 2020 IEEE 9th International Power Electronics and Motion Control Conference (IPEMC2020-ECCE Asia) N2 - A transformer with variable primary windings for wide input voltage range DC/DC converters is presented. This transformer structure allows three states of the primary winding to be converted by only two individual windings. Based on the evaluations of a two-switch flyback converter, the potential of DC/DC converters with variable structures, the implementation of such a transformer and the advantages and disadvantages are shown and discussed. Real measurement results demonstrate the noticeable increase in efficiency and the reduction in component stress, having a positive influence on the operation of the converter and the component selection during the design process. KW - DC/DC converter ; efficiency ; tapped transformer ; wide input voltage range Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1109/ipemc-ecceasia48364.2020.9367869 PB - IEEE ER - TY - JOUR A1 - Weiland, J. A1 - Luber, M. A1 - Seewald, R. A1 - Schiebahn, A. A1 - Engelbrecht, R. A1 - Reisgen, U. T1 - Structural Health Monitoring of Adhesively Bonded Joints: Proposing a new Method by use of Polymer Optical Fibers JF - Procedia Structural Integrity N2 - According to the current state of the art, adhesively bonded joints cannot be tested completely non-destructively. This is the main motivation to permanently monitor adhesively bonded joints in order to record their structural integrity and - as a result - to ensure safe load transmission. By the use of permanent non-destructive Structural Health Monitoring (SHM) methods, structural damages in the adhesive bond can be detected at an early stage. This paper presents a new method for monitoring the structural integrity of adhesively bonded joints by integrating a Polymer Optical Fiber (POF) into the adhesive layer. The sensor concept is based on a deformation transfer from the adhesive to the POF, which results in a change in the cross-sectional shape of the POF. The cross-sectional shape influences the optical light propagation in the POF, which can be detected by simple optical measuring devices. The paper describes the general principle and first results of investigations on a structural adhesive (3M Scotch-Weld DP 609). The sensor signal is in a good relation to the stress state of the adhesive layer. KW - adhesive bonding structural health monitoring polymer optical fiber single lap joint Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1016/j.prostr.2020.11.106 SN - 2452-3216 VL - 28 SP - 1249 EP - 1257 PB - Elsevier BV ER - TY - JOUR A1 - Heenemann, Maria A1 - Millet, Marie-Mathilde A1 - Girgsdies, Frank A1 - Eichelbaum, Maik A1 - Risse, Thomas A1 - Schlögl, Robert A1 - Jones, Travis A1 - Frei, Elias T1 - The Mechanism of Interfacial CO2 Activation on Al Doped Cu/ZnO JF - ACS Catalysis N2 - We report on a combined quantitative charge carrier and catalytic activity analysis of Cu/ZnO(:Al) model catalysts. The promoting effect of Al3+ on the ZnO support for CO2 activation via the reverse water–gas-shift reaction has been investigated. The contact-free and operando microwave Hall Effect technique is applied to measure charge carriers in Cu/ZnO(:Al) based model catalysts under reverse water–gas shift reaction conditions. This method allows us to monitor the electrical conductivity, charge carrier mobility, and absolute number of charge carriers. An increase in charge carrier concentration with increasing Al3+ content and its direct correlation with the catalytic activity for CO formation is found. We conclude that the increased availability of charge carriers plays a key role in CO2 activation and CO formation, which finds additional support in a concurrent decrease of the apparent activation energy and increase in the reaction order of CO2. In combination with comprehensive DFT calculations, the impact of the interfacial charge transfer, coupled to oxygen defect sites in ZnO and CO2 adsorption properties, is elucidated and highlighted. In conclusion, the results from this operando investigation combined with DFT calculations demonstrate the importance of charge transfer processes as decisive descriptors for understanding and explaining catalytic properties. KW - CO2 activation Al doped Cu/ZnO charge carrier quantification Cu−ZnO interface rWGS activity Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1021/acscatal.0c00574 SN - 2155-5435 VL - 10 IS - 10 SP - 5672 EP - 5680 PB - American Chemical Society (ACS) ER - TY - CHAP A1 - Schindler, Tobias A1 - Dietz, Armin T1 - Real-Time Inference of Neural Networks on FPGAs for Motor Control Applications T2 - 2020 10th International Electric Drives Production Conference (EDPC) N2 - Machine learning algorithms are increasingly used in industrial applications for a multitude of use-cases. However, using them in control tasks is a challenge due to real-time requirements and limited resources. In this paper, an implementation scheme for real-time inference of multilayer perceptron (MLP) neural networks on FPGAs is proposed. Design constraints for using MLPs in reinforcement learning agents for motor control applications are derived and accounted for in the implementation. Two MLP architectures are evaluated on an FPGA, and the timing and resource-usage data are reported. The real-time capability of the implementation for motor control applications is investigated for standard control frequencies. It is shown by experimental validation that real-time interference with an area-efficient implementation for motor control applications is achievable. Therefore, the proposed implementation scheme can be applied to deep reinforcement learning controllers with hard real-time requirements. Y1 - 2020 U6 - https://doi.org/https://doi.org/10.1109/edpc51184.2020.9388185 SP - 1 EP - 6 PB - IEEE ER - TY - CHAP A1 - Dernbach, Beatrice T1 - Ernährung als Medienthema BT - Einordnung und Herausforderungen des Foodjournalismus T2 - Ernährungskommunikation Y1 - 2020 SN - 9783658273156 U6 - https://doi.org/http://dx.doi.org/10.1007/978-3-658-27315-6_15-1 SP - 1 EP - 14 PB - Springer Fachmedien Wiesbaden CY - Wiesbaden ER - TY - CHAP A1 - Gerstner, Michael A1 - Maerz, Martin A1 - Dietz, Armin T1 - Design of a Wide Input Voltage Range Current-Fed DC/DC Converter Within a Reduced Duty-Cycle Range T2 - 2020 22nd European Conference on Power Electronics and Applications (EPE'20 ECCE Europe) N2 - A design of a current-fed push-pull converter for wide input voltage ranges within a reduced duty-cycle range is presented. Based on the converter conversion ratios of three different clamping configurations, the potential of the input clamping is shown analytical, simulative and by measurements. A differentiation is made between the four possible operating modes, which can be distinguished in the control of the semiconductors and the magnetomotive force in the reactor. By optimizing the turn ratios of both transformers, the wide input voltage range capability of the input clamped current-fed push-pull converter was improved, which has been proved by measurements on a real prototype. KW - DC power supply ; Design ; Converter circuit ; Measurement ; Power Supply Y1 - 2020 U6 - https://doi.org/https://doi.org/10.23919/epe20ecceeurope43536.2020.9215825 PB - IEEE ER - TY - JOUR A1 - Weiland, J. A1 - Luber, M. A1 - Seewald, R. A1 - Schiebahn, A. A1 - Engelbrecht, R. A1 - Reisgen, U. T1 - Structural Health Monitoring of Adhesively Bonded Joints: Proposing a new Method by use of Polymer Optical Fibers JF - Procedia Structural Integrity N2 - According to the current state of the art, adhesively bonded joints cannot be tested completely non-destructively. This is the main motivation to permanently monitor adhesively bonded joints in order to record their structural integrity and - as a result - to ensure safe load transmission. By the us e of permanent non-destructive Structural Health Monitoring (SHM) methods, structural damages in the adhesive bond can be detected at an early stage. This paper presents a new method for monitoring the structural integrity of adhesively bonded joints by integrating a Polymer Optical Fiber (POF) into the adhesive layer. The sensor concept is based on a deformation transfer from the adhesive to the POF, which results in a change in the cross-sectional shape of the POF. The cross-sectional shape influences the optical light propagation in the POF, which can be detected by simple optical measuring devices. The paper describes the general principle and first results of investigations on a structural adhesive (3M Scotch-Weld DP 609). The sensor signal is in a good relation to the stress state of the adhesive layer. Y1 - 2020 U6 - https://doi.org/http://dx.doi.org/10.1016/j.prostr.2020.11.106 SN - 2452-3216 VL - 28 SP - 1249 EP - 1257 PB - Elsevier BV ER -