Chair: Ruck Thawonmas, Unknown
Andrzej Cichocki, The Institute of Chemical and Physical Research (RIKEN) (Japan)
Ruck Thawonmas, The Institute of Chemical and Physical Research (RIKEN) (Japan)
Shun-ichi Amari, The Institute of Chemical and Physical Research (RIKEN) (Japan)
Two alternative neural-network methods are presented which both extract independent source signals one-by-one from a linear mixture of sources when the number of mixed signals is equal to or larger than the number of sources. Both methods exploit the previously extracted source signals as a priori knowledge so as to prevent the same signals from being extracted several times. One method employs a deflation technique which eliminates from the mixture the already extracted signals and another uses a hierarchical neural network which avoids duplicate extraction of source signals by inhibitory synapses between units. Extensive computer simulations confirm the validity and high performance of our methods.
Uwe Marschner, Dresden University of Technology (Germany)
Wolf-Joachim Fischer, Dresden University of Technology (Germany)
In this paper the mapping of a nonlinear parameter estimator onto a DSP-architecture under use of an Electronic Design Automation (EDA) tool is presented. The work concentrates on the implementation of static nonlinear process models with SISO-structure and block processing. It is shown that the synthesized processor is able to perform such a parameter estimation in less than 30 us at a reasonable chip size.
Ola Markusson, Royal Institute of Technology (KTH) (Sweden)
Torsten Bohlin, Royal Institute of Technology (KTH) (Sweden)
This contribution is a study of a method for identification of nonlinear stochastic models. Models generating electroencephalograms (EEG), based on neurophysiological knowledge axe studied, . A model-based analysis of single evoked potentials is also suggested. The main idea behind the identification is to use an inverted model, since no general predictor is available for nonlinear models. A maximum likelihood (ML) method is used to estimate the structure and the parameters of the model. To utilize a priori knowledge a 'grey-box' approach is taken.