Effective Music Genre Classification using Late Fusion Convolutional Neural Network with Multiple Spectral Features

Music genre classification is getting more and more attention amid the growing content consumption for music. Music Information Retrieval researchers have proposed various structures based on Convolutional Neural Networks that mainly achieve state-of-the-art results in the music genre classification tasks. Using multiple musical features as model inputs can improve classification accuracy. Therefore, this study proposes a new Convolutional Neural Network model using three musical features for music genre classification: Short-Time Fourier Transform, Mel-Spectrogram, and Mel-Frequency Cepstral Coefficient.