Automated Mood Categorization of Indian Melodies Using Random Forest Approach

Abstract

Most frequently, we decide to play a song or piece of music that best suits our current state of mind. Despite this significant link, the majority of the music software programmes available today do not yet offer the capability of mood-aware play-list building. By associating songs with the appropriate emotion category they communicate, music listeners might avoid spending more time manually compiling a list of tracks that suit a given mood or occasion. The challenge is to automatically and intelligently recognise this element because it is time-consuming to manually add lyrics to music with the appropriate emotion. Machine learning & data mining techniques have made significant contributions to analysing and determining how music and emotion relate to one another, which has given rise to a lot of interest in the research of mood detection in the field of music in recent years. By attempting to develop a system for automatically identifying the mood underlying the audio tracks by mining their spectral, temporal audio properties, we carry the same inspiration forward and make a contribution. Hindi songs that are popular in India are the main focus. In order to educate, prepare, and evaluate the model that captures the emotions of these audio songs, several data classification techniques were evaluated and created an open source platform for it. By using the grouping (ensemble) of random-forest-approach tested on a set of 2700 audio samples and determined the mood underlying Indian popular music with an adequate precision of 70% to 75%.