Acoustic Detection of Species Using Matching Pursuit

Abstract

Classification of bioacoustical data (audio) has become increasingly important during the last years, along with its many applications on the biology field. Traditionally, the biologists and field experts were in charge of manually categorizing audio with species annotations. With advances in data capture, the quantity of data available is overwhelming to be manually classified and categorized. Thus, biologists started using intelligent classification algorithms to automatically categorize audio with species annotations. Automatic species classification utilizes computer resources to replace the time consuming work of manually categorizing audio data. In addition, the automatic approach can be used to categorize large audio archives or real-time audio data, hence enabling the possibility of automatic retrieval. The goal of this thesis is to propose an algorithm for automatic species classification.