Speech and Audio Signal Processing: Processing and Perception of Speech and Music, Second Edition

CHAPTER 8

PATTERN CLASSIFICATION

8.1 INTRODUCTION

Much as the discipline of digital signal processing forms the technological basis for the processing of audio signals, pattern recognition is the basic field of study that underlies application areas such as speech recognition. In this chapter and the one that follows, we present some major concepts that are essential to understanding pattern-recognition technology. We begin in this chapter with a brief survey of the basic methods of classifying simple patterns. We distinguish here between pattern recognition and pattern classification by restricting the latter to the distinction between categories for sets of observed data samples or vectors. For the more inclusive term of pattern recognition we can also mean the recognition of sequences (which may be multidimensional) that are not presegmented into the patterns to be classified. In such cases it may not even be known exactly which training examples are synchronous with the class labels. For instance, in handwriting recognition, we wish to recognize the sequence of written marks with a sequence of characters (or words), but we may not know exactly where one letter ends and another begins. Similarly, in speech recognition, we wish to associate the input speech patterns with a sequence of words. Each of these examples requires the association of an observation sequence with a labeling ...