One-Class Support-Vector Machines for the Classification of Bioacoustic Time Series

Support Vector Machines (SVM) have become a widespread method in machine learning applications. In this paper we studied the one-class SVM method, whose goal is to describe the data from a single class by a set of support vectors. One-class SVMs can be used to construct multiple classifier systems (MCS) utilising the individual one-class data descriptions together with strategies to combine the classifier decisions and to resolve classifier conflict situations when a new unseen pattern has to be classified. Here the one-class SVM approach has been applied to a classification problem appearing in bioacoustic monitoring, where the species of a singing insect has to be determined.