TY - JOUR
T1 - Feature representation
T2 - Both components of the Fourier transform vs. Hartley transform
AU - Gelman, L.
PY - 2002/5
Y1 - 2002/5
N2 - A new feature representation approach was proposed [1]for those cases when one or multi-dimensional Fourier trans-forms (FT) are used for pattern recognition. This approachconsists of using simultaneously two new recognition fea-tures: real and imaginary components of the FT. It wasshown [1,2] that this approach is more generic than powerspectral density (PSD) and phase spectrum approaches andprovides better recognition e5ectiveness than the PSD ap-proach, which is a particular case of the proposed approach.This is in contrast to most recognition applications, wherePSD and phase spectrum are used.
AB - A new feature representation approach was proposed [1]for those cases when one or multi-dimensional Fourier trans-forms (FT) are used for pattern recognition. This approachconsists of using simultaneously two new recognition fea-tures: real and imaginary components of the FT. It wasshown [1,2] that this approach is more generic than powerspectral density (PSD) and phase spectrum approaches andprovides better recognition e5ectiveness than the PSD ap-proach, which is a particular case of the proposed approach.This is in contrast to most recognition applications, wherePSD and phase spectrum are used.
KW - Gaussian signal recognition
KW - Hartley transform
KW - Likelihood ratio
KW - Real and imaginary Fourier components
KW - Statistical pattern recognition
UR - http://www.scopus.com/inward/record.url?scp=0036568172&partnerID=8YFLogxK
U2 - 10.1016/S0031-3203(01)00220-5
DO - 10.1016/S0031-3203(01)00220-5
M3 - Article
AN - SCOPUS:0036568172
VL - 35
SP - 1191
EP - 1192
JO - Pattern Recognition
JF - Pattern Recognition
SN - 0031-3203
IS - 5
ER -