Audio Processing: Music Genre Classification

Our team dipped our hands into the scene of audio processing in this project. We were exposed to a wide range of audio pre-processing features. These consisted of basic features like the spectrogram which depicts the frequency and amplitude of a particular sound clip, to more complex features like the chroma feature, which depicts the total amount of different musical tones in a sound clip.

​

Multiple models were built with the aim to improve current existing ones. Other than individual models like CNN, SVM and Naive Bayes, we also explored several ensemble methods like XGBoost, CatBoost and LightGBMClasifier to help us in categorizing audio samples into their respective genres.