Neural network pattern recognition by means of differential absorption Mueller matrix spectroscopy.

Artificial neural network systems were built for detecting amino acids, sugars, and other solid organic matter by pattern recognition of their polarized light scattering signatures in the form of a Mueller matrix. Backward-error propagation and adaptive gradient descent methods perform network training. The product of the training is a weight matrix that, when applied as a filter, discerns the presence of the analytes on the basis of their cued susceptive Mueller matrix difference elements. This filter function can be implemented as a software or a hardware module to a future differential absorption Mueller matrix spectrometer.