TU Berlin

Quality and Usability LabFriedemann Köster

Inhalt des Dokuments

zur Navigation

Dr.-Ing. Friedemann Köster

Lupe

Research Field
- Quality of Speech and Audio
- Speech- and Signal Processing 

Research Topics
- Multidimensional Diagnostic Analysis of Conversational Speech Quality

Biography
Friedemann Köster studied Industrial Engineering at the Christian-Albrechts-University of Kiel. In 2012 he finished his studies and received his Dipl.-Wirtsch.-Ing. degree. Currently he is working as a research assistant at the Quality and Usability Lab of Deutsche Telekom Laboratories, TU-Berlin, where he also works towards a Phd degree in the field of quality.

Projects

Teaching

Speech Communication

Address:
Quality and Usability Lab
Deutsche Telekom Laboratories
TU Berlin
Ernst-Reuter-Platz 7
D-10587 Berlin, Germany

Telefon:

+49 30 8353 58255
+49 151 14745202

Publications

Standards for diagnosing the quality of transmitted speech and their improvements
Zitatschlüssel koester2014e
Autor Köster, Friedemann and Möller, Sebastian
Buchtitel Globecom 2014 Workshop - Telecommunications Standards - From Research to Standards
Jahr 2014
ISBN 978-1-4799-7701-7
Workshop yes
Ort Austin, TX, USA
Monat dec
Verlag IEEE Globecom 2014 Proceedings
Zusammenfassung This contribution deals with the diagnosis of the quality of transmitted speech. The two current work items approaching this topic of Study Group 12 of the International Telecommunication Union (ITU) are introduced. The first, called P.AMD, is proposed to estimate perceptual quality dimension scores and is in its final steps to become a standard. The second, called P.TCA, identifies technical causes of the transmission channel and recommends a standardized expert annotation methodology which still needs to be validated and improved. For this improvement, this paper presents a systematic approach to process and validate a set of listening material that is suitable for the P.TCA methodology. Motivation for the approach presented here is that former studies showed that the current draft explanations of degradations, which expert listeners are supposed to use for annotation, need improvement. In fact, it was found that it would be best to provide exemplarily listening material together with the P.TCA instructions. For that purpose, this paper describes the processing steps to generate the required example material and presents the results of a validation experiment in which experts annotated the processed speech material according to the P.TCA instructions. The analysis of those annotations reveals that 71\% of the example material was labelled correctly. Furthermore, the results suggest to investigate if disagreement of experts can be reduced by modifying the P.TCA annotation procedure as such, or by providing further appropriate example material.
Link zur Publikation Download Bibtex Eintrag

Navigation

Direktzugang

Schnellnavigation zur Seite über Nummerneingabe