Hs. Yang et Pr. Griffiths, Encoding FT-IR spectra in a hopfield network and its application to compound identification in open-path FT IR measurements, ANALYT CHEM, 71(16), 1999, pp. 3356-3364
Vapor-phase Fourier transform infrared (FT-IR) reference spectra of five al
cohols (methanol, ethanol, 1-propanol, 2-propanol, and 1-butanol) from 1277
to 891 cm(-1) were encoded as prototype vectors for a Hopfield network, al
ong with four baseline spectra in the same spectral region, For each input
spectrum, the Hopfield network is designed to identify one component of a m
ixture at a time (usually the major component), Inclusion of the baseline s
pectra in the Hopfield network allowed the network to classify spectra as u
nknowns, when they were not stored as prototype vectors in the network, A m
ethod that can minimize the number of spurious points and maximize the doma
in of attraction of the designed equilibrium points was used to design the
Hopfield network, Over 100 spectra of different compounds were input to thi
s network, and only three of them were incorrectly classified, The network
could usually identify each of the five alcohols correctly even in the pres
ence of noise and interfering compounds. When random noise was added to eac
h spectrum up to a level of 20% of the maximum absorbance of the analyte, t
he network could still identify each alcohol with 100% accuracy. The networ
k could also correctly identify one component of representative mixture spe
ctra, The Hopfield network was successfully used to identify ethanol and 1-
butanol from their open-path FTIR (OP/FT-IR) spectra despite the large degr
ee of similarity of these spectra, the weakness of the hands, and the fact
that the measured spectra were dominated by lines due to the absorption of
atmospheric water vapor.