Improved detection of biological substances using a hybrid neural network and infrared absorption spectroscopy

Considers basic problems that are associated with the detection of any biological substance in complex aqueous solutions using infrared absorption spectroscopy: (1) the intrinsic high background absorption of water, (2) the large number of overlapping IR absorption peaks of other molecules, and (3) the degradation of the signal of interest due to noise (usually caused by the sensing instrument itself and interference from other molecules). As a means to overcome these problems, a robust artificial neural network (ANN) detection method has been developed, which processes infrared absorption spectral data and provides a concentration decision for the dissolved substance of interest. The ANN is a hybrid structure consisting of a feedforward perceptron and a counterpropagation architecture