Publication

Abstract : Music, as all other art forms, has been used primarily as a vehicle for conveying ideas, experiences and emotions in a stylistic manner. It thus makes sense to attempt to categorize a library of music into either its style or the emotions expressed in the tracks. In this work, preliminary results of the signal processing module and machine learning module with four songs in detail and with a database of 100 songs is carried out. The signal processing algorithms employed are Mel Frequency Cepstral Coefficients and beat Histogram. Human emotions were classified based on Thayers model into Happy, Sad, Angry and Relaxed. The Machine Learning classification algorithms employed are Decision Tree Classifier and Random Forest Classifier. A low accuracy suggests improvement in the features and better machine learning algorithm before porting to Android for development as a Mobile App.