Distributed knowledge-based spectral processing and classification system for instruction and learning

The paper develops a distributed knowledge-based spectral processing and classification system which functions in one of two modes, executive and assistant. In the executive mode the system functions as a stand-alone system, automatically performing all the tasks from spectral enhancement, feature extraction and selection, to spectral classification and interpretation using the optimally feasible algorithms. In the assistant mode the system leads the user through the entire spectral processing and classification process, allowing st user to select appropriate parameters, their weights, knowledge organization method and a classification algorithm. Thus, the latter mode can also be used for teaching and instruction. It is shown how novice users can select a set of parameters, adjust their weights, and examine the classification process. Since different classifiers have various underlying assumptions, provisions have been made to control these assumptions, allowing users to select the parameters individually and combined, and providing facilities to visualize the interrelationship among the parameters.