Signal Processing for Intelligent Automation

Overview

In this chapter, we discuss basic principles governing signal processing in intelligent automation. These may appear to be somewhat academic, but their practical utility is considerable in ensuring effective and economical processing. The first principle concerns the nature of information and its definition; how it is provided in signals of various kinds (particularly images), and how processing may be configured to economize on the quantity of information to be processed and stored, with the ultimate benefit of minimizing hardware costs and processing times. The second principle concerns the nature of signals, their representation in various alternative forms, transforms to move reversibly from one form to another, and ways of exploiting the various forms to facilitate particular tasks. The third principle involves noise, a form of signal containing unwanted (and generally random) information, which masks the message information within a signal which is wanted and useful. We describe the origins and properties of noise, and show how its deleterious effects may be minimized. Finally, we consider decision making for instrumentation systems, and particularly the design of the decision process to minimize the harmful consequences of errors which are inevitable in a noisy process. Detection of the presence of a message in noise is covered as a prime application. This forms in turn an introduction to the statistical pattern recognition considered in Chapter 3.