M. Sc. Tobias HübschenRoom D-016
Kaiserstraße 2, 24143 Kiel, Germany
Phone: +49 431 880-6129
Telefax: +49 431 880-6128
Research: Instrumental Quality Assessment for Speech Communication Systems
Users of speech communication systems expect a certain quality regarding the received speech signals. The so-called Quality of Experience (QoE) expresses to what extend the users’ quality expectations are met by a specific communication service and/or a service provider. The common numerical descriptor of the QoE is the Mean Opinion Score (MOS), which is, essentially, an averaged user rating obtained from a number of subjective listening experiments. The MOS scale ranges from 1 (bad) to 5 (excellent).
To circumvent the necessity for subjective listening experiments, which inconveniently require human test subjects, instrumental quality measures such as the Perceptual Evaluation of Speech Quality (PESQ) and the Perceptual Objective Listening Quality Assessment (POLQA) have been standardized by the ITU-T. These instrumental measures estimate the outcome (MOS) of subjective listening experiments by means of signal processing. While PESQ and POLQA offer a high correlation between subjective and objective MOS, they do not offer much information about the nature or cause of the perceived quality impairments.
The Technical Cause Analysis (P.TCA) work item of the ITU-T, however, provides a more differentiated evaluation scheme based on an (ideally) exhaustive list of audible signal impairments. The subjective annotation scheme asks test subjects to select impairments based on their prominence in the current degraded speech sample. Regarding steady-state noise impairments, available annotations include e.g. pink noise, white noise, hiss, and motorboating. Hence, this annotation scheme provides information about the nature, and to some extend even the cause, of the perceived impairments.
This research aims at the development of an instrumental quality measure which estimates the subjective annotations in accordance with the P.TCA scheme. Due to the assumed interdependence of the signal impairments and the root causes (e.g. packet loss, noise reduction, bandwidth restrictions, …) within a communication system, an additional estimator for the root causes is to be developed. Such an estimator may aid service providers with debugging and optimizing their networks and services towards an improved QoE.
- speech analysis
- speech quality
- pattern recognition
- adaptive filters
- echo cancellation
|2017 - current||Research Assistant at Christian-Albrechts-Universität zu Kiel, Kiel, Germany|
|2016 - 2017||Academic Assistant at RWTH Aachen University, Germany|
|2015 - 2016||Internship at Sivantos GmbH in Erlangen, Germany|
|2014 - 2016||M.Sc. in Electrical Engineering, Information Technology, and Computer Engineering at RWTH Aachen University, Germany|
|2011 - 2014||B.Sc. in Electrical Engineering, Information Technology, and Computer Engineering at RWTH Aachen University, Germany|
T. Hübschen, M. Gimm, B. Kaulen, G. Mittag, S. Möller, G. Schmidt: Echtzeit-Rahmenwerk zur Unterstützung der Evaluierung von Sprachkommunikationssystemen, Proc. DAGA 2018 (online access)