Title :
Neural realization of fast linear filters
Author :
Jacymirski, Michal ; Szczepaniak, Piotr S.
Author_Institution :
Inst. of Comput. Sci., Tech. Univ. Lodz, Poland
Abstract :
The paper presents a neural network architecture for linear filtering of signals based on a scheme for fast calculation of orthogonal transformations. Compared to the known standard filtering schemes the network offers high training efficiency accompanied by substantial simplicity of realization. Additionally, the same basic network topology enables realization of diverse transformations.
Keywords :
filtering theory; learning (artificial intelligence); matrix algebra; network topology; neural net architecture; fast linear filters; linear filtering; network topology; neural network architecture; orthogonal transformations; signal filtering; training efficiency; Computer architecture; Computer science; Discrete Fourier transforms; Filtering; Image coding; Maximum likelihood detection; Network topology; Neural networks; Neurons; Nonlinear filters;
Conference_Titel :
Video/Image Processing and Multimedia Communications 4th EURASIP-IEEE Region 8 International Symposium on VIPromCom
Print_ISBN :
953-7044-01-7
DOI :
10.1109/VIPROM.2002.1026646
