TU Berlin

Quality and Usability Labkolloquium_WS10/11

Inhalt des Dokuments

zur Navigation



Mondays, starting March 22, 2010


TEL Auditorium 1 and 2 (20th floor), Ernst-Reuter-Platz 7, 10587 Berlin

This research colloquium is a weekly event with various invited speakers. It is open to anyone who is interested in the general area of usability and human-computer interaction. Researchers in this area will present overviews of their work. The colloquium is organized by Deutsche Telekom Laboratories. If you have any questions, please contact Hamed Ketabdar, Shiva Sundaram or Sebastian Möller

Please contact us if you want your email to be added to our colloquium mailing list.


Colloquium Talks

Please notice: The talk takes place in room 1118/19, 11th floor, Ernst-Reuter-Platz 7

Posthuman Designs: The Accumulation of Biocultural Capital

SPEAKER: Prof. Andy Miah

Over the last 10 years, the growth of human enhancement technologies has been accompanied by global debates about the promise and peril of an emerging posthuman era. This presentation will discuss recent contributions from bioartists and biodesigners to the ensuing bioethical debate about transhumanism, which promises to reconstitute how we make sense of biological and social norms. In so doing, it introduces the concept of biocultural capital to explain how the pursuit of better humans is consistent with the pursuit of a healthy, long life.

Professor Andy Miah, BA, MPhil, PhD, is Chair of Ethics and Emerging Technologies in the Faculty of Business & Creative Industries at the University of the West of Scotland, a Global Director for the Centre for Policy and Emerging Technologies, Fellow of the Institute for Ethics and Emerging Technologies, USA and Fellow at FACT, the Foundation for Art and Creative Technology, UK. www.andymiah.net He is author of ‘Genetically Modified Athletes’ (2004), co-author of ‘The Medicalization of Cyberspace (2008) and Editor of ‘Human Futures: Art in an Age of Uncertainty’ (2008).

HOST: Tom Bieling
Time: 9:30-10:30
Room: Auditorium II, Ernst-Reuter-Platz 7

Feature Enhancement for Reverberant Speech Recognition

SPEAKER: Prof. Dr. Reinhold Häb-Umbach

We present a model based feature enhancement approach to the automatic recognition reverberated speech. The estimation of the dereverberated feature vector sequence is treated as a Bayesian estimation problem, where the goal is to infer the posterior density of the clean feature vector, given the current and past observed reverberant features. As an priori model, which captures the dynamics of the nonreverberant features, we employ switching linear dynamical models, which have already been successfully employed in noisy speech recognition. Further we derive a novel observation model which links the observed reverberant to the nonreverberant features. Bayesian inference is then carried out using the Interacting Multiple Model algorithm. We conducted experiments on the AURORA V reverberant digit recognition task and observed a reduction of word error rate by up to 75% compared to the recognition of unprocessed reverberant speech. The obtained results are among the best obtained on that database.

Reinhold Haeb-Umbach (M'89) received the Dipl.-Ing. and Dr.-Ing. degree in Electrical Engineering from RWTH Aachen University in 1983 and 1988, respectively. From 1988 to 1989 he was a postdoctoral fellow at the IBM Almaden Research Center, San Jose, CA, conducting research on coding and signal processing for recording channels.
From 1990 to 2001 he was with Philips research working on various aspects of automatic speech recognition. Since 2001 he is a full professor in communications engineering at the university of Paderborn, Germany.
His main research interests are in statistical speech signal processing and recognition.

HOST: Sebastian Möller
Time: 14:00 - 16:00
Room: Auditorium I, Ernst-Reuter-Platz 7

TITLE: Exploring the Design of novel systems that bring creative hands-on practices in digital environments: A case study in Animation

SPEAKER: Mariza Dima

HOST: Christine Kühnel

How do we design a digital workspace that encompasses the tacit skills of creative practicioners by succesfully connecting their bodily responses to the place in which they are manifested?
In this talk, I will reflect on the above question by presenting and discussing the development of a digital system for character Animation which is built upon key elements of the traditional practice of Stop-Motion Animation.
I will describe our exploration of expressive and direct interaction through the employment of physical interfaces and creative mappings between the user's gesture and the desired output.
By conducting a series of design key studies, each one aimed at reconceptualising the Stop-Motion animator's workspace, we endevoured to investigate what is lost, gained or transformed by adding digitality to a conventionally 'hands-on' practice.
The presentation will evolve around the design process and evaluation of the system and move on to discuss the issues encountered and the extend to which the system forms a link between the traditional and the Computer Generated practice of Animation.

Mariza Dima is a digital designer currently working towards a PhD in Human Computer interaction at the University of Edinburgh. Her research focuses on exploring the design process of mediating physical skilled hand practices in digital workspaces through the employment of gestural interfaces.Her work has been published in various papers and consortia including British HCI 2009. She holds a BSc in Applied Mathematics completed at the National Technical University of Athens in 2005, followed by an MSc in Design and Digital Media at the University of Edinburgh. In the majority of her work she is interested in creating interactive environments which blend the physical and the digital space having expressivity, and immediacy as design driving forces.
Please notice the exceptional date and time:
The talk takes place between 14:30-16:00, in Auditorium 1, Ernst-Reuter-Platz 7

Data quality in the Entity Name System

SPEAKER: Oleksiy Chayka

In this work we show how the generic paradigm of quality of data (“fitness for use”) can be implemented for assessment of data quality in an Entity Name System (ENS), which aim is management of global unique identifiers of entities from the real world. We demonstrate an application of a well-known methodology for data quality assessment (AIMQ) to such data retrieval and management systems as the ENS. As a result we define a set of quality dimensions that we use for the quality assessment and management in our system. We show how such set of quality dimensions produces quality requirements for input data from users of the ENS.
We also propose an approach for evaluation of algorithms for measurement of one of quality dimensions, namely data staleness. Space requirements of those algorithms are evaluated against accuracy of their prediction of data staleness. The evaluation approach is tested by implementation of various prediction algorithms run on Wikipedia’s data, which simulates in our case various update patterns of web data.

Oleksiy Chayka received his BSc in Artificial Intelligence from the Kharkov National University of Radioelectronics (Ukraine). He received joint MSc degree from the University of Trento (Italy) and RWTH Aachen (Germany) in 2007. In his current PhD course at the University of Trento his main research area is data quality management (data quality definition, assessment, analysis and improvement) which he is pursuing in European project OKKAM and recently was working on in MASTER project. He is also interested in system analysis and application of data mining techniques to data quality assessment.

HOST: Dmitry Butenkov



Schnellnavigation zur Seite über Nummerneingabe