Computer Analysis and Classification of Waveforms and Pictures. Part I. Waveforms

摘要

Fundamental operations used in processing waveforms and pictures are collected together under the framework of the Bayes approach. Algorithms corresponding to the operations are stored in a computer. An operator experimentally determines a sequence of operations which maximizes a performance criterion related to the accuracy with which the waveforms or pictures can be correctly classified. Two classes of operations are considered. One class contains operations which are interpretive allowing the operator to perform data (waveforms or pictures) analysis. These interpretive operations provide an interactive capability between the operator and the computer. The second class of operations is for actual data processing with the end result of discrimination (includes detection and classification). The operations in this second class are based on the philosophy that a priori knowledge is available and is to be utilized. Within this second class of operations are three subclasses of operations. (a) Feature extraction--utilization of a priori knowledge through a model specific to the problem; (b) Dimensionality reduction using estimation; (c) Discrimination (supervised and unsupervised) using estimation. (Author)