Classification of active sonar echoes using a one-class classification technique
Abstract
A typical approach to data classification based on machine learning algorithms is binary classification. This in-volves the classifier to be trained using representative data sets provided from two object classes. In reality, da-ta from one of the classes may be not well-defined or readily available and so the one-class classification tech-nique is gaining popularity. In this research we apply this method to the problem of classification using active sonar echoes from different classes of objects. A one-class classification research tool was developed in Matlab® to implement several one-class classification techniques found in literature. The tool was applied to three sets of data: simulated, laboratory and at-sea. The performance of the selected classifiers on different da-ta sets will be discussed in this paper.