17th International Symposium on
Mathematical Theory of Networks and Systems
Kyoto International Conference Hall, Kyoto, Japan, July 24-28, 2006

MTNS 2006 Paper Abstract

Close

Paper MoP09.4

Nitta, Masuhiro (Nara Inst. of Science and Tech.), Sugimoto, Kenji (Nara Inst. of Science and Tech.)

ICA Based Blind Identification Via Exact Parameterization

Scheduled for presentation during the Regular Session "Linear system identification I" (MoP09), Monday, July 24, 2006, 16:35−17:00, Room J

17th International Symposium on Mathematical Theory of Networks and Systems, July 24-28, 2006, Kyoto, Japan

This information is tentative and subject to change. Compiled on March 28, 2024

Keywords Linear system identification, Statistical learning, Multidimensional systems

Abstract

This paper addresses a blind identification problem of multi-input-multi-output discrete-time systems based on independent component analysis. Assuming that input signals are independent, the paper proposes a method which makes it possible to identity a transfer matrix only from the observation signals by maximizing independence of estimated source signals. In order to treat a vector ARMA model with exact parameterization, a feedback-structured learning rule is proposed in this paper. Numerical simulations are carried out to illustrate the effectiveness of the proposed algorithm.