Signal Processing for Cognitive Radios

8

BAYESIAN NONPARAMETRIC CLASSIFICATION

8.1 INTRODUCTION

Spectral activity identification problem can be either a classification or a clustering problem. Once a wideband cognitive radio has detected a set of spectral activities located at certain frequencies, determining the type of each signal requires a classification stage as shown in Figure 2.4. Suppose that we are given a set of observations . For simplicity, let us assume that each . Each y_i corresponds to a specific class θ_i. The general classification problem can then be stated as

Classification approaches can be broadly divided into two classes: supervised and unsupervised. In supervised classification, there are initial training samples that can be used as a guide in classifying the given data . For example, suppose that the training data were given where denotes the class of the training data sample . Note that the distinct ...