forum | IKS - Details

forum | IKS: Vortrag von Patrick A. Naylor

Imperial College, London, UK
Multichannel Blind Acoustic System Identification with Under‐modelling
Montag, 9. Mai 2016
17:00 Uhr
Hörsaal 4G


Das Institut für Kommunikationssysteme startete im Sommersemester eine neue Veranstaltungsreihe unter dem Titel "forum | IKS".

Die Veranstaltungsreihe setzt sich mit aktuellen Themen aus dem Bereich der digitalen Kommunikationssysteme auseinander und dient gleichzeitig als Kommunikationsplattform für den regen Austausch zwischen Hochschule und Industrie in einem informellen Umfeld. Das Programm richtet sich an Studierende, an die Vereinsmitglieder der Gesellschaft für Kommunikationstechnik e.V., an Angehörige der Fakultät sowie an alle weiteren fachlich Interessierten.

Im Anschluss an das Fachprogramm hatten die Gäste die Möglichkeit, bei einem Imbiss mit dem Vortragenden zu diskutieren. 

Being able to identify acoustic systems that represent the propagation from a sound source to multiple sensors provides valuable information about the acoustic environment. This information can be used beneficially in several acoustic signal processing tasks including for example direction-of-arrival estimation and dereverberation. In typical scenarios, the identification process does not have access to the sound source’s signal and therefore must operate blindly. An existing class of approaches involves minimization of the cross-relation error and exploits the knowledge that all microphone signals originate from a single source. Cross-relation minimization can successfully identify multichannel acoustic systems, subject to certain identifiability conditions, when there is no noise and the channel lengths are known and of low order (typically below 128 coefficients, for example). These conditions are not normal satisfied in practice since acoustic systems in reverberant conditions can be several thousand coefficients in length at commonly used sampling frequencies and, in the limit, infinite. Furthermore, noise is always present. In this talk, the task of multichannel acoustic system identification involving under-modelling is addressed. In particular, a method is presented for blind identification of only the direct path and early reflections of a multichannel acoustic system. Robustness is aided by the inclusion of a sparsity assumption. The performance is compared with several cross-relation-based approaches from the literature for varying levels of under-modelling.

Patrick Naylor received his BEng degree in Electronic and Electrical Engineering from the University of Sheffield, U.K., and the PhD. degree from Imperial College, London, U.K. Since 1990 he has been a member of academic staff in the Department of Electrical and Electronic Engineering at Imperial College London. His research interests are in the areas of speech, audio and acoustic signal processing. He has worked in particular on adaptive signal processing for dereverberation, blind multichannel system identification and equalization, acoustic echo control, speech quality estimation and classification, single and multi-channel speech enhancement and speech production modelling with particular focus on the analysis of the voice source signal. In addition to his academic research, he enjoys several fruitful links with industry in the UK, USA and in mainland Europe. He is the Chair of the IEEE Signal Processing Society Technical Committee on Audio and Acoustic Signal Processing, a director of the European Association for Signal Processing (EURASIP) and formerly an associate editor of IEEE Signal Processing Letters and IEEE Transactions on Audio Speech and Language Processing.