Signal Processing and Speech Communication Laboratory
homeevents › PhD defense Michael Soudan

PhD defense Michael Soudan

Start date/time
Mon Dec 5 13:00:00 2011
End date/time
Mon Dec 5 14:30:00 2011
Seminarraum D3.07 am Institut für Computer Graphik und Wissensvisualisierung, Inffeldgasse 16c/II, 8010 Graz

Chairman: Ao.Univ.-Prof.Dr. A. Hofer
Examiner: Univ.-Prof.Dr. G. Kubin
Examiner: Prof.Dr. P. Händel (KTH Royal Institute of Technology, Stockholm)

Low Complexity Correction Structures for Time-Varying Systems

Time-varying systems are encountered in various fields of engineering. If the time-varying behavior of a system is undesired, it produces a distorted output signal. Dedicated timevarying systems can be cascaded with the original system to correct the impact of the undesired time-varying behavior on the output signal. In applications where a high reconstruction accuracy is important, the computational cost of designing and employing flexible digital correction systems remains challenging. In particular, the computational load becomes a major challenge if the digital correction system needs to be redesigned for each time instant. In this thesis, low complexity correction methods for the design of linear time-varying correction filters are presented. These filters can be applied to postcorrect or precorrect linear time-varying systems. In order to mitigate the computational complexity of the filter design, a low complexity filter design algorithm for the least-squares norm is derived which exploits prior knowledge for the design and can be applied to postcorrect or precorrect linear time-varying systems. Furthermore, correction structures for weakly time-varying linear systems are presented. Employing these structures provides a significant reduction of the computational effort, that is required to redesign the correction system, at the expense of an increased complexity of the correction structure. Finally, the extension of the latter concept to time-varying Volterra systems is explored.