Chair of
Multimedia Communications and Signal Processing
Prof. Dr.-Ing. André Kaup

System identification

Field of activity: Audio and Acoustic Signal Processing
Research topic: Signal Improvement and Detection
Staff: Prof. Dr.-Ing. Walter Kellermann
Dipl.-Ing. Christian Hofmann
PhD Shmulik Markovich-Golan

System identification is a "classical" problem of analog and digital signal processing, and describes the task of identifying an existing unknown physical or technical system by an appropriate model. The fundamental scenario is illustrated by Fig. 1.

Systemidentifikation

Fig. 1: Fundamental task of system identification


Typically, the estimation of the true coefficients and parameters of the system (h) is obtained by applying a parallel model (ha) and continuous analysis of the input and output data (x and ya), such that the residual error between the true and the estimated system (e) vanishes. For the sake of complexity and robustness, it is often assumed that (h) can be modelled sufficiently well by a purely linear transversal model. However, in practice the output of the unknown system cannot be observed directly but is impaired by background noise and/or local interferers. These additional signals hamper the identification and are usually modelled as additive noise (n). Moreover, since the underlying system has to be modelled as (more or less) time-variant, the adjustment of the filter (ha) is often performed by an iterative adaptation algorithm, where the optimization typically aims at minimization of a squared error criterion (Least-Mean-Square, LMS).

Besides the conventional estimation of the filter coefficients, determining the optimal filter parameters, as, e.g., order and size of the employed filter memory, is also desirable in many applications. In this way, the risk of under- or overmodelling is reduced and the adaptive structures are implemented only with the necessary computational complexity. At the Chair of Multimedia Communications and Signal Processing, we have therefore developed fully adaptive FIR filter structures with dynamically "growing" or "shrinking" memory size. Using this methodology, current research focuses on extending these algorithms towards a more generalized identification of linear and nonlinear sytems.

For many applications, the considered system (h) represents a multiple-input/multiple-output (MIMO) system. Suitable adaptation algorithms for those scenarios require a large computational effort and typical multichannel loudspeaker signals may not allow for a unique system identification. Models for acoustic MIMO systems in the wave domain can alleviate these problems and were already be applied to system identification in different applications.

Areas of application

Since system identification can essentially be considered as the common goal of most applications exploiting optimum linear filtering [1], it is a key element of numerous modern signal processing algorithms. This is supported by the fact that the identified systems are usually required for subsequent processing stages as, e.g., compensation of signal distortions or calculation of the system's inverse.

In the LMS audio group, system identification challenges are tackled in the following projects:


[1] S. Haykin, Adaptive Filter Theory, Prentice Hall, Upper Saddle River (NJ), USA, 2002 (4th Edition).

Publications

2016-71
CRIS
M. Schneider, W. Kellermann
   [link]   [doi]   [bib]

Multichannel Acoustic Echo Cancellation in the Wave Domain With Increased Robustness to Nonuniqueness
IEEE Transactions on Audio, Speech and Language Processing (IEEE TASLP) Vol. 24, Online Publication, Num. 3, Pages: 518 - 529 , Mar. 2016
2016-68 M. Schneider, W. Kellermann
   [bib]

APPARATUS AND METHOD FOR LISTENING ROOM EQUALIZATION USING A SCALABLE FILTERING STRUCTURE IN THE WAVE DOMAIN
US 9,338,576 B2, Num. US 9,338,576 B2, May 2016
2016-67 M. Schneider, W. Kellermann
   [bib]

APPARATUS AND METHOD FOR LISTENING ROOM EQUALIZATION USING A SCALABLE FILTERING STRUCTURE IN THE WAVE DOMAIN
EP 2 754 307 B1, Num. EP 2 754 307 B1, Europe, Aug. 2016
2016-65 M. Schneider, W. Kellermann
   [bib]

APPARATUS AND METHOD FOR PROVIDING A LOUDSPEAKER-ENCLOSURE-MICROPHONE SYSTEM DESCRIPTION
EP 2878138, Num. 2878138, Oct. 2016
2016-64 M. Schneider, W. Kellermann
   [bib]

APPARATUS AND METHOD FOR ROVIDING A LOUDSPEAKER-ENCLOSURE-­MICROPHONE SYSTEM DESCRIPTION
US 9,326,055 , Num. 9,326,055 , USA, Apr. 2016
2016-63 M. Schneider, W. Kellermann
   [bib]

Apparatus and Method for providing a Loudspeaker-Enclosure-Microphone System Description
JPA 6038312, Num. 6038312, Japan, Nov. 2016
2016-32
CRIS
M. Rupp, W. Kellermann, A. Zoubir, G. Schmidt
   [link]   [doi]   [bib]

Advances in adaptive filtering theory and applications to acoustic and speech signal processing (Editorial)
EURASIP Journal on Advances in Signal Processing (JASP) Vol. 2016:63, Online Publication, May 2016
2016-3
CRIS
C. Hofmann, W. Kellermann
   [bib]

Source-Specific System Identification
IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Pages: 161--165, Shanghai, China, Mar. 2016
2015-43
CRIS
C. Hümmer, R. Maas, C. Hofmann, W. Kellermann
   [doi]   [bib]

A Bayesian network approach to linear and nonlinear acoustic echo cancellation
EURASIP Journal on Advances in Signal Processing (JASP) Online Publication, Num. 2015:98, Pages: 1--11, Nov. 2015
2015-39
CRIS
S. Meier, W. Kellermann
   [bib]

Analysis of the Performance and Limitations of ICA-Based Relative Impulse Response Identification
European Signal Processing Conf. (EUSIPCO), Pages: 414--418, Nice, France, Aug. 2015
2014-39
CRIS
K. Reindl, S. Meier, H. Barfuss, W. Kellermann
   [link]   [doi]   [bib]

Minimum Mutual Information-Based Linearly Constrained Broadband Signal Extraction
IEEE Transactions on Audio, Speech and Language Processing (IEEE TASLP) Vol. 22, Num. 6, Pages: 1096-1108, Jun. 2014
2014-22
CRIS
R. Maas, C. Hümmer, A. Schwarz, C. Hofmann, W. Kellermann
   [bib]

A Bayesian Network View on Linear and Nonlinear Acoustic Echo Cancellation
IEEE China Summit and International Conference on Signal and Information Processing (ChinaSIP), Pages: 495--499, Xi'an, China, Jul. 2014
2014-14
CRIS
C. Hofmann, C. Hümmer, W. Kellermann
   [bib]

Significance-Aware Hammerstein Group Models for Nonlinear Acoustic Echo Cancellation
IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Pages: 5934--5938, Florence, Italy, May 2014
2014-11
CRIS
C. Hümmer, C. Hofmann, R. Maas, A. Schwarz, W. Kellermann
   [bib]

The elitist particle filter based on evolutionary strategies as novel approach for nonlinear acoustic echo cancellation
IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Pages: 1315--1319, Florence, Italy, May 2014
2013-60
CRIS
L.A. Azpicueta-Ruiz, M. Zeller, A.R. Figueiras-Vidal, W. Kellermann, J. Arenas-Garcia
   [link]   [doi]   [bib]

Enhanced adaptive volterra filtering by automatic attenuation of memory regions and its application to acoustic echo cancellation
IEEE Transactions on Signal Processing (IEEE TSP) Vol. 61, Online Publication, Num. 11, Pages: 2745--2750, 2013
2013-58
CRIS
M. Zeller
   [bib]

Generalized Nonlinear System Identification using Adaptive Volterra Filters with Evolutionary Kernels
Dr. Hut Verlag, München, 2013
2013-51
CRIS
K. Reindl, S. Markovich-Golan, H. Barfuss, S. Gannot, W. Kellermann
   [doi]   [bib]

Geometrically Constrained TRINICON-based Relative Transfer Function Estimation in Underdetermined Scenarios
IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), Pages: 1-4, New Paltz, NY, USA, Oct. 2013
2013-34
CRIS
M. Schneider, W. Kellermann
   [doi]   [bib]

Large-Scale Multiple Input/Multiple Output System Identification in Room Acoustics
21st International Congress on Acoustics , Pages: 1--9, Montreal, Canada, Jun. 2013