ABSTRACT
Infrared spectral comparisons are frequently employed for molecular structure elucidation. A wiggle parameter was employed to compensate for band center coding inconsistencies. Algorithms employed to sort library spectra by similarity to the unknown spectrum are known as search metrics. The most popular methods employed for pattern recognition are the linear learning machine and K-nearest-neighbor methods. The linear learning machine is an iterative process in which a multidimensional plane is sought that is capable of separating pattern classes. The goal of pattern recognition is to develop algorithms that are capable of recognizing the presence of molecular structure features based solely on spectral information. Vectors containing structure-specific information often cluster in one area of pattern space. An expert system attempts to mimic the actions of an experienced spectroscopist in making spectrum/structure correlations. Expert systems consist of a knowledge base, inference engine, and a user interface.
