Music Genre Classification of Audio Signals

Music genre classification is an interesting but challenging task given its inherent subjective nature. Through this project we treat genre classification of raw audio signals as a machine learning problem. We extensively discuss relevant feature extraction necessary for classification and various machine learning techniques applied to it. Based on the dataset selected we discuss the accuracy of classification of each algorithm applied and also give relevant insights. We obtained the best classification accuracy of 66.73% for a 10 genre classification problem using adaboost with Support vector machines as the weak learner

Report - Download

Presentation - Download

Matlab Code - Github