Acoustic scene classification: A comprehensive survey

Abstract

Acoustic scene classification (ASC) has gained significant interest recently due to its diverse applications. Various audio signal processing and machine learning methods have been proposed for ASC. The volume and scope of ASC publications covering theories, algorithms, and applications have also been expanded. However, no recent comprehensive surveys exist to collect and organize the knowledge, impeding the ability of researchers and its applications. To fill this gap, we present an up-to-date overview of ASC methods, covering earlier works and recent advances. In this work, we first define a general framework for ASC, starting with a historical review of previous research in the ASC field. Then, we review core techniques for ASC that have achieved good performance. Focus on machine learning based ASC systems, this work summarizes and groups the existing techniques in terms of data processing, feature acquisition, and modeling. Furthermore, we provide a summary of the available resources for ASC research and analyze ASC tasks in Detection and Classification of Acoustic Scenes and Events (DCASE) challenges. Finally, we discuss limitations of the current ASC algorithms and open challenges to possible future developments toward practical applications of ASC systems.