New methods for anechoic demixing with application to shift invariant feature extraction [Elektronische Ressource] / Lars Omlor

Ulm UniversityFaculty of Engineering and Computer ScienceNew methods for anechoicdemixing with application toshift invariant feature extractionLars Omloraus St. WendelDissertationzur Erlangung des Doktorgrades Dr.rer.nat.der Fakultat fur Ingenieurswissenschaften und Informatikder Universitat Ulmsupervised byProf. Martin A. GieseSection for Computational SensomotoricsUniversity TubingenandProf. Heiko NeumannInstitute of Neural Information ProcessingUlm UniversityAmtierender Dekan: Prof. Dr.-Ing. Michael WeberGutachter: Prof. Dr. Heiko Neumannhter: Dr. Martin A. Giese(Gutachter:) Prof. Dr. Andreas SchillingTag der Promotion: 09:03:2010AbstractBlind source separation problems emerge in many applications, where signals can bemodeled as superpositions of multiple sources. Many popular applications of blindsource separation are based on linear instantaneous mixture models. If speci c invari-ance properties are known about the sources, e.g. translation or rotation invariance,the simple linear model can be extended by inclusion of the corresponding transforma-tions. When the sources are invariant against translations (i.e. spatial displacementsor time shifts) the resulting model is called anechoic mixing model.The main focus of this thesis is the development of new mathematical framework forthe solution of the anechoic mixing problem and the successive derivation of concretealgorithms.