A fully-differential CMOS implementation of Oja´s learning rule in a dual-synapse neuron for extracting principal components for face recognition

Author

Spencer, Ronald G. ; Sanchez-Sinencio, Edgar

Author_Institution

Analog & Mixed-Signal Center, Texas A&M Univ., College Station, TX, USA

Volume

2

fYear

1999

fDate

1999

Firstpage

1102

Abstract

A fully-differential, CMOS implementation of a self-organizing, dual-synapse neuron with on-chip learning for real-time facial feature extraction is presented. The adaptation of the network follows Oja´s learning rule and the synaptic weight vector is shown to adapt to the principal component vector of the set of two-dimensional input vectors

Keywords

CMOS analogue integrated circuits; SPICE; analogue multipliers; face recognition; feature extraction; learning (artificial intelligence); neural chips; operational amplifiers; principal component analysis; self-organising feature maps; signal flow graphs; Gilbert cell multiplier; HSPICE; OTA-C integrator; Oja´s learning rule; autocorrelation matrix; eigenvectors; face recognition; fully-differential CMOS implementation; on-chip learning; principal components extraction; real-time facial feature extraction; self-organizing dual-synapse neuron; signal flow graphs; synaptic weight vector; two-dimensional input vectors; Autocorrelation; Data mining; Face recognition; Facial features; Flow graphs; Network synthesis; Neurons; Principal component analysis; Signal synthesis; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1999. 42nd Midwest Symposium on

Conference_Location

Las Cruces, NM

Print_ISBN

0-7803-5491-5

Type

conf

DOI

10.1109/MWSCAS.1999.867829

Filename

867829