Unsegmented Data

The unsegmented case is the simplest classifier form where the complete input data ${\bf x}$ is presented directly to the feature transformation, the feature ${\bf z}$ is extracted, and presented as a single vector to the joint-PDF estimation, typically a kernel-based approach such as Gaussian mixtures (Section 13.2). Applications include problems where the data consists of a fixed-size measurement, such as in text categorization based on a specified set of word-counts.