- Kommunikationstechnik-Kolloquium - lists.rwth-aachen.de

Communication Technology Colloquium at IKS

by Simone Sedgwick

Dear subscribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Wednesday, January 18, 2022* *Speaker*: Konstantin Wehmeyer *Time:* 2:00 p.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Lecture: ***Deep Learning-Based Speech Synthesis as Post-Processing of a Noise Reduction /Audio and speech signals are often disturbed by noise signals in frequency- and/or time-limited parts. To attenuate or remove these distortions, several methods, including deep learning- based approaches, are known. Often, however, only the magnitude spectrum is processed and the phase spectrum is taken over unchanged due to its comparatively lower relevance. Consequently, the noisy phase is reused when synthesizing the waveform from the processed magnitude spectrum. Therefore, distortions in the magnitude spectrum can be reduced, but not in the phase spectrum which inevitably leads to a deterioration in speech quality and intelligibility./ /This thesis presents methods that allow a reconstruction of the phase spectrum of speech signals based on noise-reduced magnitude spectra. At the Institute of Communication Systems at RWTH Aachen University a phase reconstruction algorithm was developed and this algorithm has already been evaluated in a previous study for the case of smoothed magnitude spectra. It was shown that the deep neural network (DNN) used can benefit from targeted training on the smoothed magnitude spectra even without further modification of the network structures. However, even slight smearing of the magnitude spectra already leads to a significant loss in performance compared to the use of perfect magnitude spectra. In this work, therefore, the DNNs used are optimized for the case of noise-reduced magnitude spectra. // / /Several deep learning-based models are introduced and compared with each other and with the models already developed. Their properties and aspects such as causality are addressed. Moreover, a new loss function and assessment measure specifically designed to estimate and assess the phase spectrum of speech signals is developed and tested. In order to be able to evaluate the results as independently as possible of a specific type of noise reduction, ideal masks are developed, used, and discussed. / All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Simone Sedgwick Institute of Communication Systems(IKS) Prof. Dr.-Ing. Peter Jax RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26956(phone) +49 241 80 22254(fax) sedgwick(a)iks.rwth-aachen.de https://www.iks.rwth-aachen.de/

1 month

1
3
0 0

Communication Technology Collquium at IKS

by Irina Esser

Dear subscribers of the Colloquium Newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Wednesday, 30. November 2022** **Speaker:* Frederick Pietschmann *Time*: 14:00 *Location:* Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture*: Perceptual Optimization and Evaluation of a Binaural Signal Modification Algorithm With individualized binaural signals, it is possible to reproduce auditory scenes such that the signal is perceived similar to the real scene. However, perceptual similarity is no longer achieved when the binaural signal doesn’t fully adapt to diﬀerent listeners and diﬀerent orientations of the listener’s head. To address these problems, a perceptually motivated algorithm referred to as the Binaural Cue Adaptation (BCA) system has been developed at the Institute of Communication Systems. The BCA system is capable of adding both interactivity and individualization to existing binaural signals, thereby achieving a higher degree of perceptual similarity to a corresponding real auditory scene. In this thesis, a perceptual optimization of the existing BCA system is conducted in that new approaches for some components of the algorithm are proposed, all parametrization options are identiﬁed and the overall best parametrization is chosen. To identify the best parametrization, both an isolated analysis of individual components is conducted and a perceptually motivated optimization procedure for a full system analysis is proposed and implemented. Finally, a perceptual evaluation based on the result of the perceptual optimization is realized. For this, two listening tests with a total number of 17 participants are conducted – one for a normal and one for a highly reverberant scenario. The results of these listening tests suggest that signals produced by the optimized BCA system achieve a high degree of perceptual plausibility for both reverberation scenarios, with an averaged 2AFC probability to detect a BCA-generated signal of 0.563 for the normal scenario and 0.604 for the highly reverberant scenario. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be fount at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) esser(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

6 months

1
0
0 0

Kommunikationstechnisches Kolloquium am IKS

by Irina Esser

Sehr geehrte Abonnenten des Kolloquium-Newsletters. gerne informieren wir Sie über den nächsten Termin unseres Kommunikationstechnischen Kolloquiums. *Mittwoch, 23. November 2022* *Vortragender: *Philipp Tigges *Zeit: *11:00 Uhr *O**rt*: hybrid - Hörsaal 4G und https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Vortrag*: Charakterisierung der Körperschallübertragung von Sprache für Hearables Körperschall ist ein wesentlicher Bestandteil bei der Wahrnehmung der eigenen Stimme. Das Übertragungsverhalten von Körperschall ist deutlich komplexer als bei Luftschall, da verschiedenste Gewebearten mit unterschiedlichen Eigenschaften durchlaufen werden. Da die Erzeugung von Sprachlauten ein vielschichtiger Prozess ist und die Bildung eines spezifischen Klangs auch mit dem Erzeugungsort zusammenhängt, liegt die Annahme eines phonemspezifischen Übertragungsverhaltens nahe. Ziel der Arbeit ist es, die Übertragung von Sprache über Körperschall zu charakterisieren, um die Eigenstimmwahrnehmung bei der Benutzung von Hearables zu verbessern. Für eine Auswahl an Phonemen wurde der Körperschall per Beschleunigungssensor und der Schall im inneren des Ohrkanal über ein Mikrofon gemessen. Die beiden Signale wurden dann auf lineare Zusammenhänge hin untersucht. Auf den Signalen wurden dann Filter ausgelegt, um aus dem Signal des Sensors das Signal im Ohrkanal zu schätzen und dann den Okklusionseffekt zu reduzieren. Es hat sich gezeigt, dass zwischen dem Signal des Beschleunigungssensors und dem Signal im Ohrkanal ein linearer Zusammenhang gefunden werden kann. Das Übertragungsverhalten der verschiedenen Phoneme unterscheidet sich, jedoch sind die Unterschiede klein genug, so dass sich Filter zur Okklusionsreduktion auch auf andere Laute anwenden lassen. Auch für gesprochenen Text können Filter gefunden werden, die den Okklusionseffekt effektiv reduzieren können. Zusätzlich wurden auch einige Geräusche, die sich auch über Körperschall ausbreiten, analysiert. Hier war es deutlich schwerer lineare Zusammenhänge zu finden. Im weiteren Verlauf wurden dann die Rahmenbedingungen realistischer gestaltet, indem der Sekundärpfad mit einbezogen wurde. Außerdem fordern reale Anwendungen kausalen Filtern. Diese zeigten sich ähnlich performant, wie ihre akausalen Äquivalente, jedoch mit starken Überhöhungen vor allem zu hohen Frequenzen, da sie nah an der Kausalitätsgrenze arbeiten müssen. Im letzten Teil der Analyse konnten erste Einblicke in die Frage, ob die erstellten Filter eine ausgeprägte Abhängigkeit von der jeweiligen Person haben, gewonnen werden. Diese Abhängigkeit hat sich insbesondere bei den auf einzelnen Phonemen ausgelegten Filtern gezeigt. Alle Interessierten sind herzlich eingeladen, eine Anmeldung ist nicht erforderlich. Allgemeine Informationen zum Kolloquium sowie eine aktuelle Liste der Termine des Kommunikationstechnischen Kolloquiums finden Sie unter: http://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) ronkartz(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

6 months, 1 week

1
0
0 0

Kommunikationstechnisches Kolloquium am IKS

by Irina Esser

Sehr geehrte Abonnenten des Kolloquium-Newsletters, gerne informieren wir Sie über den nächsten Termin unseres Kommunikationstechnischen Kolloquiums. *Mittwoch, 16. November 2022* *Vortragende:* Jana Lorenz *Zeit*: 09:00 Uhr *Ort*: hybrid - Hörsaal 4G und https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Vortrag:* Vergleich adaptiver Algorithmen für die aktive Störgeräuschkompensation bei Kopfhörern in komplexen Schallfeldern Lärmbelästigung ist ein alltägliches Problem. Es gibt viele Ansätze diese zu reduzieren. Kopfhörer mit aktiver Störgeräuschunterdrückung (ANC, engl. Active Noise Cancellation) sind einer davon. Sie kompensieren vor allem tieffrequente Geräusche, welche passiv nur unzureichend gedämpft werden können. Dabei hat das Störgeräusch einen großen Einfluss auf die aktiven Kompensationsmöglichkeiten. Dieses kann sich in realen Umgebungen bezogen auf seine Einfallsrichtung oder seine statistischen Eigenschaften verändern und somit die aktive Störgeräuschunterdrückung erschweren. Daher sind adaptive Feedforward-Algorithmen von großer Bedeutung, da diese sich kontinuierlich an das sich verändernde Störgeräusch anpassen können. So werden in dieser Arbeit zwei ANC-Filterstrukturen, der Filtered-x Least Mean Square Algorithmus (FxLMS) und der Adaptive Linear Combiner (ALC), betrachtet. Es wird untersucht, wie sich die feste adaptiv gewichtete Parallelfilterstruktur des ALC, verglichen zum FxLMS-Algorithmus, auf das Konvergenzverhalten und die aktive Dämpfung auswirkt. In dieser Arbeit werden beide Algorithmen anhand von Schallfeldern verschiedener Komplexität verglichen. Es werden ruhende, sich räumlich bewegende und komplexere Schallfelder verwendet. Dabei erreicht die ALC-Filterstruktur, verglichen zum FxLMS-Algorithmus, auf Kosten einer geringeren maximal möglichen aktiven Dämpfung, eine schnellere Konvergenzzeit. Alle Interessierten sind herzlich eingeladen, eine Anmeldung ist nicht erforderlich. Allgemeine Informationen zum Kolloquium, sowie eine aktuelle Liste der Termine des Kommunikationstechnischen Kolloquiums finden Sie unter: http://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) esser(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

6 months, 2 weeks

1
0
0 0

Kommunikationstechnisches Kolloquium am IKS

by Irina Esser

Sehr geehrte Abonnenten des Kolloquium-Newsletters, gerne informieren wir Sie über die nächsten Termine unseres Kommunikationstechnischen Kolloquiums. *Mittwoch, 9. November 2022* *Vortragender:* Nils Lattasch *Zeit*: 14:00 Uhr *Ort:* hybrid - Hörsaal 4G und https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Vortrag*: Koordinatentranformation für die Adaptive Aktive Störgeräuschunterdrückung in Kopfhörern Bei der aktiven Störgeräuschkompensation wird versucht, ein Störsignal mit einem destruktiv interferierenden Signal auszulöschen. Dazu wird in dieser Arbeit die feedforward Topologie verwendet. Für diese kann gezeigt werden, dass das optimale Filter ein IIR Filter impliziert. Da die Störgeräusche in der Praxis facettenreich sind, ist es ebenfalls sinnvoll adaptive Lösungen zu verwenden. In etablierten adaptiven Verfahren wie dem FxLMS Algorithmus kann versucht werden, ein FIR Filter mit möglichst vielen Filterkoeffizienten zu erstellen, um eine lange Impulsantwort zu generieren. Gleichzeitig ergibt sich jedoch ein Problem in der Adaption, da Systeme in vielen Variablen, in diesem Fall viele Filterkoeffizienten, langsam adaptiert werden können. Eine mögliche Lösung des Problems stellt eine Koordinatentransformation dar, bei der zeitinvariante IIR Filter, Teil eines adaptiven Gesamtsystems sind. Somit kann die Anzahl der Filterkoeffizienten reduziert werden und zusätzlich besitzt das sich ergebende Filter eine unendliche Impulsantwort. Ziel dieser Arbeit ist es eine Koordinatentransformation für die adaptive aktive Störgeräuschkompensation zu realisieren. Die Koordinatentransformation stellt durch eine frei wählbare Polstelle einen neuen Freiheitsgrad im Design eines ANC-Systems dar. In dieser Arbeit wird die Performance und der Ressourcenverbrauch der Koordinatentransformation systematisch untersucht. Zur Implementierung wird das sogenannte Laguerre-Netzwerk verwendet. Daraus ergeben sich neue Adaptionsvorschriften für das adaptive System, welche in dieser Arbeit hergeleitet werden. Zudem wird untersucht, welche Polstellen in dem Anwendungsfall ANC vorteilhaft sind. Des Weiteren wird untersucht, wie eine Konvergenzgrenze bezüglich der maximalen Schrittweite abgeschätzt werden kann. Abschließende Experimente erfolgen auf Basis von Messdaten am Beispiel eines gängigen in-ear Kopfhörers. Es zeigt sich, dass die Performance sensitiv gegenüber der Wahl der Polstelle und Schrittweite ist, bei geeigneter Wahl jedoch die Anzahl der Filterkoeffizienten um ein vielfaches reduziert werden können. und *Mittwoch, 9. November 2022* *Vortragender:* Christian Wolf *Zeit:* 15:0 Uhr *Ort*: hybrid - Hörsaal 4G und https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Vortrag*: Auslegung von Übersprechkompensationssystemen mittels Verfahren der robusten Regelung Das Abspielen von Binauralsignalen über Lautsprecher ist, im Gegensatz zur Verwendung von Kopfhörern, durch akustisches Übersprechen nachteilig beeinflusst. Das Übersprechen kann jedoch durch geeignete Vorfilterung des Binauralsignals so modifiziert werden, das es effektiv unterdrückt wird. Dies ist das grundsätzliche Prinzip der Übersprechkompensation. Das Vorfilter wird dabei als Übersprechkompensationsfilter bezeichnet. Eine gängige Methode zur Auslegung eines Übersprechkompensationsfilters ist die Methode der kleinsten Quadrate im Zeitbereich. Diese Arbeit stellt ein Vorgehen vor, mit dem Übersprechkompensationsfilter mithilfe von Verfahren aus der robusten Regelung, wie zum Beispiel der H2- oder H∞-Synthese, entworfen werden können. Zunächst wird gezeigt, wie sich die Übersprechkompensation als Problem der robusten Regelung modellieren lässt, sodass Lösungsverfahren wie die H2- oder H∞-Synthese auf das Problem angewendet werden können. Das allgemeine Problem wird um wählbare Gewichtungsfunktionen erweitert, sodass die Performance des Systems gezielt und frequenzselektiv beeinflusst werden kann. Zudem wird eine Regularisierung bezüglich der Amplitudengänge des Übersprechkompensationsfilters vorgesehen, mit der die Verstärkung des Übersprechkompensationsfilters auf ein in der Praxis sinnvolles Maß reduziert werden kann. Weiterhin wird untersucht, welchen Einfluss die Latenz des akustischen Systems auf die Performance hat und wie dies im Design berücksichtigt werden sollte. Im Anschluss daran wird der erarbeitete Ansatz mit der H2- und H∞-Synthese mit der Methode der kleinsten Quadrate im Zeitbereich verglichen. Durch eine theoretische Herleitung und eine Simulation wird gezeigt, dass die H2-Synthese und die Methode der kleinsten Quadrate im Zeitbereich direkt zusammenhängen. Insbesondere scheint sich die Methode der kleinsten Quadrate im Zeitbereich für große Längen des Übersprechkompensationsfilters den Ergebnissen mit der H2-Synthese anzunähern. Ein weiterer praxisrelevanter Aspekt, der in der Arbeit genauer beleuchtet wird, ist die Unsicherheit im akustischen System und deren Einfluss auf die Performance. Allgemein ist die Unsicherheit in praktischen Systemen verschiedenen Ursachen geschuldet. In dieser Arbeit wird die Unsicherheit durch die Variation von Kopfdrehung und der Geometrie verschiedener Zuhörer genauer untersucht. Es wird gezeigt, dass die Performance im Allgemeinen nicht robust gegenüber den untersuchten Variationen ist, jedoch bei tiefen Frequenzen weniger signifikant als bei hohen Frequenzen. Alle Interessierten sind herzlich eingeladen, eine Anmeldung ist nicht erforderlich. Allgemeine Informationen zum Kolloquium, sowie eine aktuelle Liste der Termine des Kommunikationstechnischen Kolloquiums finden Sie unter: http://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) ronkartz(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

6 months, 3 weeks

1
0
0 0

Communication Technology Colloquium at IKS

by Irina Esser

Dear subscribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Wednesday, October 12, 2022* *Speaker*: Christian Prick *Time:* 10:00 a.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Lecture*: Adaptive Active Noise Cancellation for Headphones using Virtual Sensing In recent years, active noise canceling headphones became widespread because they can attenuate low-frequency noise and therefore complement the passive dampening characteristics of headphones. In order to further improve their performance, methods to increase the attenuation at higher frequencies are of interest. As the main limiting factor for high frequency attenuation is the transfer characteristics from the error microphone to the eardrum, virtual sensing algorithms have to be used. These require a knowledge of the acoustic paths to the eardrum, which are influenced by variations of different kinds. This thesis investigates the influences of the direction of arrival (DoA) and fit on the virtual sensing paths and proposes methods for their approximation. Furthermore, a combined method is proposed, which takes both the fit and DoA into account to approximate the virtual sensing paths at runtime. The resulting performance improvements are evaluated for both in- and over-ear headphones using dummy head measurements and acoustic simulations. The results indicate that the influence of the DoA and fit depend on the type of the headphones. The approximations thus achieve different performance improvements, with the fit generally having a larger impact. For both types of headphones, the combined method is able to improve attenuation. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) ronkartz(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

7 months, 3 weeks

1
0
0 0

Communication Technology Colloquium at IKS

by Irina Esser

Dear subcribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Friday, September 2, 2022* *Speake**r:* Vitor Horst Duque *Time*: 10:00 a.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture:* Autoencoder based Waveform Enhancement for Software Defined Radio Applications Deep learning is a trend that has made it's way into communication systems and gained a lot of interest already. this presentation discusses a machine learning-based approach to add novel features to existing waveform implementations for software defined radios. Such additional features include the reduction of the peak-to-average power ratio, spread spectrum abilities and interference mitigation. The developed machine learning structures use autoencoders to add these features. The results indicate that realizing additional features by means of autoencoders is a promising concept for future waveform development. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) esser(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

9 months

1
0
0 0

Communication Technology Colloquium at IKS

by Irina Esser

Dear subscribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Monday, June 13, 2022* *Speaker*: Zihang Wei *Time*: 02:00 p.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture*: Investigations on Supervised System Identification Algorithms The system identification task aims at inferring the room impulse response of a specific acoustic enclosure. System identification is mandatory in applications such as acoustic echo cancellation and cross talk cancellation. Traditional gradient-based algorithms such as normalized least mean square algorithm uses FIR filters to estimate the RIRs, unfortunately, in a relatively high dimension. A novel dual-stage algorithm is proposed in this thesis. The algorithm performs a state update where allowed states are located on a manifold. In a first stage, an undercomplete autoencoder is trained over the RIR data set. In the second stage, we perform the system identification tasks. Here the problem is reformulated such that the latent state is updated instead of the full impulse response. The trained decoder is then exploited to transform the latent variables to a proper impulse response. Evaluation is made between the reconstructed RIR with reference to the true RIR. In this thesis, at first, the simulation framework generates RIR data set. Then autoencoders with different layer setups are trained on the generated data set. The qualified autoencoders are employed in the inference stage to perform the system identification tasks. Two crucial parameters, i. e., the latent dimension size and updating step size of the manifold are investigated under different SNR conditions. It is demonstrated that under noisy conditions, the proposed method outperforms the traditional NLMS approach. Evaluation results also show that lower bottleneck size design benefits the system identification with adverse noise and *Monday, June 13, 2022* *Speaker*: Johannes Imort *Time*: 03:00 p.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture*: Online Learning of Loudspeaker Nonlineraities for Acoustic Echo Cancellation Hands-free communication is pervasive throughout modern society, requiring a robust cancellation of the echo from the far end speech signal that is emitted from the loudspeaker to the microphone. Acoustic echo cancellation addresses this issue typically by employing linear adaptive filters. However, in reality, the echo path is nonlinear due to the non-ideal characteristics of acoustic transducers and power amplifiers operated at their physical limits. This thesis introduces and investigates a novel approach to tackle nonlinear AEC by estimating the nonlinear reference signal using a deep neural network and a differentiable Kalman filter. The hybrid system is designed to learn loudspeaker nonlinearities directly from data, enabling end-to-end training on data composed of pairs of the far end reference and near end microphone signals. In contrast to previous neural network-based solutions that have been tailored toward one particular loudspeaker, the proposed system aims to be generalizable for different loudspeaker nonlinearities. Therefore, inspired by linear adaptive filtering, the recurrent architecture explicitly takes advantage of the information in the residual echo in order to estimate the nonlinearity adaptively. The proposed approach was evaluated for both simulated and measured data. The results indicate that the architecture could enable faster convergence and better steady state performance than related adaptive approaches. Furthermore, some examples demonstrated that the performance of a method that makes use of oracle knowledge could be surpassed, evidently because the models adapt to the linear acoustic echo path, too. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) ronkartz(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

11 months, 3 weeks

1
0
0 0

Communication Technology Colloquium at IKS

by Irina Esser

Dear subscribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Wednesday, June 1, 2022* *Speaker*: Pattaratorn Santivarakom *Time:* 11:15 a.m. *Location*: hybrid - Lecture room 4G and https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Bachelor-Lecture*: Sound Source Localization Using Binaural Signals Binaural signals are similar to the sounds that humans hear with their left and right ears when the signal is recorded with a suitable device. As the human auditory system is able to localize sound sources, the directions of sound sources can also be found in binaural signals. This thesis addresses the performance of sound source localization algorithms using binaural signals. The algorithms regarded in this thesis are based on the concept of beamforming. Several algorithms from the literature are systematically compared and their basic building blocks are recombined. From this, a total of 26 variants are identified. The performance of an algorithm is determined by comparing the accuracy of the estimated source source’s directions with other algorithms. Furthermore, error metrics are suggested that take into account the varying signal powers in different time-frequency bins. According to the results, there are two algorithms that always estimate the same source source’s directions and they also provide the most accurate source source’s directions in most situations. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: https://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) esser(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

1 year

1
0
0 0

Communication Technology Colloquium at IKS

by Irina Esser

Dear subscribers of the colloquium newsletter, we are happy to inform you about the next date of our Communication Technology Colloquium. *Monday, April 25, 2022* *Speaker*: Alexander Sobolew *Time*: 03:00 p.m. *Location*: https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture*: Investigation of Specialized Recurrent Units for Acoustic Echo Cancellation In today's communication, hands-free devices e.g. remote communication are widely used. Without further action, these would suffer from an acoustic echo that arises from the coupling between speaker and microphone. To minimize these disturbances, acoustic echo cancellation is indispensable. Model-based adaptive algorithms exist to solve this issue. However, they require careful tuning of parameters whose optimum differs between devices and acoustic situations. In this thesis, a new data-driven approach for acoustic echo cancellation is developed and investigated. In contrast to the purely model-based approach, the algorithm is supposed to learn the optimal performance from data without the necessity to be tuned. In new situations, the unknown parameters should be estimated. At its core, the novel structure is similar to a frequency adaptive filter. However, it is extended by the gating mechanism known from recurrent neural networks. The development also includes the determination of optimal training paradigms. When choosing the model structure, attention is paid to a reasonable training complexity. Major challenges in this thesis include the investigation of the gating mechanism, which is represented by a learn gate and a reset gate. The prior is used to estimate a time-varying step size of the iterative algorithm. Gated Recurrent Units provide an internal memory to accommodate the sequential information in speech, while skip connections optimize the gradient flow during training. Independent use of a reset gate to reset the impulse response estimation in case of situation change is outperformed by exploitation of weight sharing. Using weight sharing, the learn and reset gates have direct information about each other's behavior due to a shared partial network. It was shown that when using backpropagation through time, the truncation order can be reduced to a certain extent, which reduced the training complexity but did not decrease the performance. The developed model outperforms the tuned Fast Block Normalized Least-Mean-Square algorithm in reconvergence speed and steady-state performance in far-end single talk and double talk. Furthermore, our model repeatedly outperforms the tuned diagonalized Kalman filter in certain scenarios and offers significantly improved overall performance in single talk. and *Monday, April 25, 2022* *Speaker*: Alexej Sobolew *Time*: 04:00 p.m. *Location*: https://rwth.zoom.us/j/97904157921?pwd=SWpsbDl0MWhrWjY1ZkZaeFRoYmErZz09 Meeting-ID: 979 0415 7921 Passwort: 481650 *Master-Lecture*: Investigation of Generative Neural Networks for Speech Enhancement Speech enhancement aims to reduce noise in speech signals and is widely used in hearing aids and mobile speech communication. Speech synthesis, on the other hand, aims to generate high-quality human speech and is used, e. g., in text-to-speech generation. Noise reduction and speech synthesis can be combined since conventional noise reduction methods often only improve the magnitude spectrum and keep the noisy phase. However, the phase has an important influence on speech quality and intelligibility. In addition, training neural networks with complex spectrograms is more difficult, so it is reasonable to first denoise the magnitude spectrum and subsequently synthesize the waveform of the speech based on it. Mentioned applications often require low execution times and low computational overhead. This is achievable by exploiting parallel processors using the non-autoregressive property and by reducing the number of parameters in the neural network. Hence, this thesis aims to investigate noise reduction, speech synthesis, and joint interaction. In this thesis, the first use case considered is phase reconstruction and speech synthesis based on clean data. A non-autoregressive three-stage speech enhancement system is developed for the second use case of combined noise reduction and speech synthesis based on magnitude spectra. For speech synthesis on clean magnitude spectra such as mel-spectrograms, the neural network called WaveGlow from Nvidia is taken as a basis. WaveGlow achieves similar subjective performance as the Griffin-Lim algorithm but is better suited for fast applications. For the reduction of parameters in WaveGlow, the SqueezeWave is used, resulting in a decrease in the number of parameters and the inference time by up to 70%. During the switch to additional noise reduction besides speech synthesis, it is shown that WaveGlow alone is not suitable for performing both tasks simultaneously. Consequently, the problem is divided into three stages: masking, inpainting, and synthesis. The models for masking and inpainting are adapted to mel-spectrograms and studied in detail for noise reduction. As result, they are able to reduce the noise significantly. It is worth noting that in this thesis the mel-filter is used as downsampling adapted to human perception due to its non-linear property, resulting in a reduction of the number of computations in the first two models. Subsequently, the performance of the entire three-stage speech enhancement system is investigated. It improves the speech quality and intelligibility of noisy data while exploiting parallel processors and can compete with existing state-of-the-art methods. The system achieves better noise reduction than the Convolutional Recurrent Network (CRN) and additionally does not rely on the noisy phase. All interested parties are cordially invited, registration is not required. General information on the colloquium, as well as a current list of dates of the Communication Technology Colloquium can be found at: http://www.iks.rwth-aachen.de/aktuelles/kolloquium -- Irina Esser Institute of Communication Systems (IKS) RWTH Aachen University Muffeter Weg 3a, 52074 Aachen, Germany +49 241 80 26958 (phone) ronkartz(a)iks.rwth-aachen.de http://www.iks.rwth-aachen.de/

1 year, 1 month

1
0
0 0

2023

2022

2021

2020

2019

2018

Kommunikationstechnik-Kolloquium