NIR calibration in non-linear systems: different PLS approaches and artificial neural networks

Application of NIR spectroscopy for firmness evaluation of peaches

“…The main characteristics hindering its expansion at first is that NIR spectra typically consist of broad, weak, non-specific, extensively overlapped bands (Blanco et al, 2000). Until multivariate calibration methods became widely available and accepted, NIR spectroscopy technique has been widely studied and applied.…”

Determination of <i>trans</i> Fat in Selected Fast Food Products and Hydrogenated Fats of India Using Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR) Spectroscopy

“…height at approximately 966 cm 1 , which is the characteristic peak of isolated trans double bond 23 , was evaluated in second derivative spectrum using Advance Origin 6.0 software. The pictures changed in most of the cases when the spectral profiles were doubling derivate.…”

Chemometrics in Process Analytical Technology (PAT)

“…Artifi cial neural networks (ANNs) [53,67,68] simulate the parallel processing capabilities of the human brain, where a series of processing units (aptly called ' neurons ' ) are used to convert input variable responses into a concentration (or property) ' output ' . As a chemometric quantitative modeling technique, ANN is rather different than all of the regression methods mentioned previously, for three main reasons: (1) the model structure is best expressed using a ' map ' or architecture, rather than a simple mathematical expression; (2) the model parameters are not determined by least squares or any other closed -form mathematical operation, but rather by a ' searching ' algorithm, which starts with a random selection of starting parameters and performs sequential alteration of the parameters ( ' training ' ) by testing with individual calibration samples; and (3) ANNs allow for nonlinear modeling through the use of nonlinear transfer functions in the model structure.…”