Название: Machine Recognition of Patterns
Автор: Ashok K. Agrawala
Издательство: IEEE Press
Год: 1977
Формат: djvu
Страниц: 469
Размер: 39,49 МБ
Язык: english
Over the years, the field of pattern recognition has attracted workers from a variety of areas such as engineering, system theory, statistics, linguistics, psychology, etc., resulting in a vast literature containing abstract mathematical approaches as well as highly pragmatic techniques. This literature is scattered in a large number of journals in several fields. At least three IEEE journals regularly publish pattern recognition papers. While several textbooks are available for a beginner in the field, a need often arises to go to the source, which, due to the nature of the literature in this field, is not always a straightforward task. This collection of selected readings is not limited to early or "classic'' papers. Rather it is designed to be a companion volume to many of the textbooks in pattern recognition and to be asource of useful references for engineers interested in developing the many potential applications of pattern recognition methodology.
Pattern recognition is an applied field which tends to discover techniques to solve practical problems. Over the years, however, a lot of theory has been developed with little application of the theory being attempted to the extent that, for a practical problem, the approach to be taken Is rarely clear. Within the last few years, there has been a tendency to use interactive systems which allow the user to apply a variety of techniques to the problem at hand. The problems of dimensionality, sample size, and the error rate often tend to limit the design goals. A highlight of this volume is the selections on these topics. Nine papers covering all the major issues involved in these areas are included.
Although pattern recognition techniques can be applied to a variety of problems in a number of fields, only optical character recognition (OCR), blood cell recognition, and isolated speech recognition have reached a stage of commercial use. However, except in the already mature area of OCR, no fully documented case histories are available in the literature. Although the work on OCR is nearly two decades old, the early approaches did not exploit the theoretical advances and relied only on ad hoc techniques. Due to proprietary reasons, most of the more recent commercial applications (e.g., blood cells) described in the literature give general discussions, often leaving out crucial details. The application papers included here are intended to give an indication of potential rather than specific case histories. One hopes that well-developed case histories will appear in the not too distant future.