Feature Engineering and Classification of Different Sound Waves

Abstract

With the rapid growth of multimedia technologies, a large number of music resources are now available online, leading to increased interest in classifying different music genres. The main objective of a music recommendation playlist is to identify a set of songs belonging to a similar genre. Machine learning, transfer learning, and deep learning concepts can be used to build a robust music classifier that can tag unlabelled music and improve the user experience of media players with music files. However, existing approaches in the past decade have several limitations, such as the manual extraction of features and traditional machine learning classification techniques, which impact the classification accuracy, particularly for multiclass classification problems and huge data sizes.