Elektronische Sprachsignalverarbeitung

 

Keynotes


Keynote Speaker: Reinhold Häb-Umbach
(Paderborn University, Germany)


Sprachtechnologien für digitale Assistenten


Abstract


Digitale Sprachassistenten, wie z.B. Amazon Alexa oder Google Home, sind High-Tech Produkte, welche eine Vielzahl von Sprachtechnologien vereinen, von der Signalverarbeitung über die Erkennung und Synthese bis hin zur Verarbeitung natürlicher Sprache. Dieser Vortrag gibt einen Überblick über diese Technologien, mit einem Fokus auf die Verfahren, die eine robuste Erkennung trotz entfernter Mikrofone ermöglichen. Dabei wird deutlich, wie eine geschickte Kombination von klassischen Signalverarbeitungskonzepten und maschinellen Lernverfahren zu sehr effizienten Lösungen führt.


Bio


Reinhold Haeb-Umbach is a professor at Paderborn University, Germany. After obtaining a PhD degree from RWTH Aachen University, he was PostDoc at the IBM Almaden Research Laboratory, and from 1990 - 2001 he worked as a Senior Scientist at Philips Research, Aachen and Eindhoven. Since 2001 he is professor of Communcations Engineering at Paderborn University. He has more than 200 scientific publications, and recently co-authored the book Robust Automatic Speech Recognition - a Bridge to Practical Applications (Academic Press, 2015). He is also the lead author of a recent IEEE Signal Processing Magazine article on digital home assistants. He is a fellow of the International Speech Communication Association (ISCA). His main research interests are in the fields of statistical signal processing and machine learning, with applications to speech enhancement, automatic speech recognition and unsupervised learning from speech and audio.



Keynote Speaker: Michael Beigl
(Karlsruhe Institute of Technology, Germany)


Human-Computer Interaction closer to the Body


Abstract


In the early days of Computing, the first Human-Computer Interaction system designers implement quite inconvenient-to-use interaction methods such as card readers. Human operators interact with these interfaces far away from the computer, resulting in a spatial distance of the interaction between user and computer. Inch by inch, every development of computing technology reduces this distance: With the advent of typewriters that succeeded card readers users sit in front of the interface, and then with the direct interaction of PCs and workstations users are in a "touch" distance to the computing system. The invention of Smartphones and Smart Watches started a new era of intimate interaction between the human (body) and the computer, with computing that is always on, always accessible and always with the user. Today the potential of such truly wearable interaction interfaces is still not fully leveraged, but the next wave of Human-Computer Interaction is already setting up. Interfaces that are attached directly to the skin of the body, put into the body or even implanted in the body.
This keynote will provide a closer look on interaction methods, interfaces and the required technology for upcoming novel types Human-Computer Interaction (HCI) that is attached on and partially in the body. Such interfaces show great potential not only to replace existing interfaces with the computer, but also to create completely new type of interfaces thus inventing novel type of Human-Computer Interaction. For example, HCI on the human body must not necessary be explicit, as this is mostly the case with traditional HCI where you e.g. press a (virtual) button to tell a computer what you want him to do. Being closer to the body, a computer system can use also implicit and subconscious expressions as an input – without requiring the user to consciously "think" about it. Even more, novel interfaces may allow implementing novel human senses or even allowing for implicit, subconscious learning.
In particular, the talk will draw a light on novel body worn Human-Computer Interfaces by showing examples of these interfaces such as Haptic Displays, Second Skin Interfaces, Tattoo Computing, Earables or Head attached interfaces. I will illustrate what the technical requirements are to implement them, how the interfaces can be attached at or in the body, and how to use these interfaces to create novel ways of Human-Computer Interaction.


Bio


Michael Beigl is professor for Pervasive Computing Systems at the Karlsruhe Institute of Technology (KIT) and head of the TECO research lab. He obtained both his M.Sc. and Ph.D at the University of Karlsruhe (now KIT). Previously he was professor at the TU Braunschweig from 2006-2010, visiting associate professor at Hide Tokuda Labs, Keio University, Japan in 2005 and research director of the TECO, University of Karlsruhe, Germany from 2001-2005. He is heading the national competence center for Big Data AI, the Smart Data Innovation Lab (SDIL) and the state competence center for Big Data AI in Baden-Württemberg, the Smart Data Solution Center (SDSC-BW) since 2014. His research interests evolve around blending human with computing, with specific interest In Human-Computer Interaction, sensing systems and blending Artificial and Human Intelligence. Contact him at Michael.Beigl@kit.edu.



Keynote Speaker: Jesko L. Verhey
(Otto von Guericke University, Germany)


Listenung through a Cochlear Implant


Abstract


A cochlear implant (CI) partially restores hearing for profoundly deaf patients. It is a surgical implanted electronic device. It includes an array of electrodes that are inserted into the inner ear. These electrodes stimulate the auditory nerve by short electric pulses. To transmit spectral information of the sound it uses the concept of frequency-place transformation, which is observed in the healthy cochlea. Thus, a key part of the signal processing is a filterbank analysis. Due to the filterbank processing and a subsequent envelope extraction, the sound processing of a CI severely affects sound information. This talk will focus on two aspects of sound perception through a CI. The first aspect is perception of speech. CI processing is optimised for speech recognition. However, some aspects of speech are not preserved when transmitted thorough a CI which causes CI listeners to have more difficulties in complex acoustic environment to understand speech than normal-hearing listeners. This talk will address the role of reverberation in speech recognition and the ability to recognize talkers through a CI. The second focus is on music perception with a CI. Music still poses a particular problem to CI users. Whereas the CI signal processing roughly preserves the temporal structure such as rhythms, recognition of melodies or certain instruments is significantly harder. This is presumably partly due to the limited frequency resolution of a CI, leading to a reduced pitch percept. In the light of the altered pitch percept with a CI we investigated perception of musical consonance. We have addressed this issue by specific experiments where we focussed on consonance perception for chords in isolation and in successions (cadences). Overall, it is shown that a CI is a powerful utility to restore hearing but that further research is required to overcome the limitations of a CI in complex acoustic environments.


Bio


Jesko Verhey is a professor at the Otto von Guericke University Magdeburg, Germany. He studied physics in Göttingen, Germany, and then moved to Oldenburg, Germany, where he did his PhD in 1998. As a research associate, he moved 2000 to Cambridge (United Kingdom). At the end of 2002, he returned to Oldenburg as Juniorprofessor in Neurosensory Sciences, did his Habilitation in Applied Physics in 2007 and was Research Professor until December 2010. In January 2011, he started at the Medical Faculty of the Otto von Guericke University Magdeburg as full Professor for Audiology. His main research interest is auditory perception. His focus is on the perception of normal-hearing and hearing impaired listeners. His research fields include test methods in audiology, listening through cochlear implants, as well as basic and applied psychoacoustics. Currently he is the President of the German Acoustical Society (DEGA).


Organised by:

Sponsored by:

By Ulrich Arendt.

ESSV 2020

By Ulrich Arendt.