chapter  1
12 Pages

Introduction

As human beings, our physiological structure enables us to look, speak, hear, taste, smell, touch, and feel our surroundings. If we look at a familiar object, say a tree, we can tell that it is a tree and not a chimney or a water reservoir. Our ability to recognize and differentiate between objects that we see, hear, and touch would not be possible without the presence of a powerful sensory system. Our brain and our nervous system allow us to experience our surroundings through a combination of senses and memories. It is estimated that the human brain comprises approximately 80 to 120

billion neurons, which respond to a multitude of actions, perceptions, and emotions. From a physical perspective, our brains could be considered as large-scale interconnected networks of sensory systems and memories. Seeing, identifying, and recalling what we have observed make up a significant portion of the activities conducted within these large-scale networks. The recall process, also known as recognition, is a part of information pro-

cessing that happens in our brain-nervous systems. Watkins and Gardiner [5] suggested a two-stage theory, in which recall begins with a search and retrieval process that is followed by a decision or recognition process. The correct information is chosen in the decision process from that which has been retrieved. Recognition of objects and other forms of events or stimuli is part of our brains’ activities. Strong interest in this area has led to further understanding of the recognition process and how it can be performed us-

ing computational approaches. The study of the recognition process based on computational theories and the biological behavior of the nervous system can be traced back to the 1950s, when digital computers started being used for information processing. The ability to recognize and extract valuable information from raw data has motivated extensive research on pattern recognition techniques. Such techniques aspire to emulate the behavior of neural systems in living organisms. To fully understand the concept of pattern recognition, there is a need to

differentiate between some of the terms that are commonly used interchangeably, namely pattern recognition, data mining, and pattern classification. Pattern recognition is the process of identifying an object or entity based on

its descriptions and a set of measurements, commonly referred to as a pattern. Keeping with the previous example, a tree can be characterized by its vertical cylindrical shape, leaves, bark, and branches. In pattern recognition, we use these features to identify and differentiate a tree from other objects, such as a chimney or water reservoir. To obtain useful information from data, it is important for applications to

extract features or patterns. Pattern extraction from data is commonly known as data mining and involves uncovering patterns, associations, anomalies, interesting data structures, and traces of events. Recognition of patterns plays an important role in data mining applications in a variety of fields, including the life and physical sciences, economics, finance, and engineering. Pattern classification is the process of assigning an object or entity to a

class that shares similar characteristics or features. For example, biological taxonomy uses pattern classification to identify and label individuals as a class of species that have similar characteristics and behaviors. The aim of any pattern recognition scheme is to achieve high recall ac-

curacy for any recognition problem. However, almost every approach has to substantially increase its algorithmic complexity to accommodate this goal. Some promising approaches in assimilating and comprehending the functionalities of biological nervous systems have been proposed. Nevertheless, the highly cohesive procedures and processing-centric algorithmic design of these approaches may limit the capabilities of such approaches. Because requirements for the intensive collection and retrieval of data are appearing as a consequence of the data deluge phenomenon, it is important that we also consider the recognition process from the perspective of scalability.