Transputer based back-propagation neural net emulation system

Author

Jou, I-Chang ; Tsai, Yung-Chang

Author_Institution

Minist. of Commun., Chung-Li, Taiwan

fYear

1990

fDate

1-3 May 1990

Firstpage

1879

Lastpage

1882

Abstract

A transputer based back-propagation neural net emulation system is presented. The authors propose one mapping approach for mesh architecture. The major challenge of this research work is to map the back-propagation neural net model onto transputer arrays for mesh architecture. The OCCAM parallel programming language is used for programming the operations of these transputers. Using this transputer based emulation system, a test case for Mandarin (Chinese voice) speech zero to nine recognition is implemented, and encouraging results, i.e. the recognition rate of inside test 99% and outside test 96%, have been obtained

Keywords

computerised pattern recognition; neural nets; parallel algorithms; parallel architectures; speech recognition; transputers; Chinese voice; Mandarin speech recognition; OCCAM; back-propagation; mapping approach; mesh architecture; neural net emulation system; parallel programming language; transputer arrays; transputer based emulation system; Artificial neural networks; Emulation; Natural languages; Neural networks; Neurons; Parallel processing; Parallel programming; Speech recognition; Switches; System testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1990., IEEE International Symposium on

Conference_Location

New Orleans, LA

Type

conf

DOI

10.1109/ISCAS.1990.112044

Filename

112044