A Survey on Audio Content-Based Classification

Abstract

With the increased usage of mobile electronic devices, social media platforms, and electronic based applications in everyday life, the upload and usage of multimedia clips is exponentially increasing every year. There are many research initiatives tackling the challenge of multimodal video classification and it has been found that audio-based classification is less computationally expensive and just as effective, in many cases. However, research targeted towards acoustic-based detection is still in its initial stages. Audio content-based classification pertains to several domains: music and speech signal processing, which are relatively popular research interests, and event, genre and scene-based classification which are still areas that need a lot of development. There is also a problem with the difficulty of assessing performances of different systems with a unified audio dataset, due to the lack of development in this field. The objective of this study is to present information about audio processing techniques and studies conducted under several classification tracks; namely acoustic-based event, genre, and scene detection, as well as combination-based classification works. ? 2017 IEEE.