Title :
Performance and implementation of invariant source separation algorithms
Author :
Cardoso, Jean-François
Author_Institution :
ENST, CNRS, Paris, France
Abstract :
This paper focuses on the equivariant nature of source separation: the unknown parameter of source separation is an invertible matrix, i.e. it belongs to a multiplicative group. In this instance, inference theory calls for `equivariant´ estimation. This paper discusses some consequences of equivariance with respect to implementation and performance of source separation algorithms
Keywords :
inference mechanisms; invariance; matrix multiplication; signal detection; equivariant estimation; equivariant nature; inference theory; invariant source separation algorithms; invertible matrix; multiplicative group; signal detection; Adaptive algorithm; Additive noise; Convergence; Inference algorithms; Neural networks; Qualifications; Signal processing; Signal processing algorithms; Source separation; Yttrium;
Conference_Titel :
Circuits and Systems, 1996. ISCAS '96., Connecting the World., 1996 IEEE International Symposium on
Conference_Location :
Atlanta, GA
Print_ISBN :
0-7803-3073-0
DOI :
10.1109/ISCAS.1996.540358
