Deep Learning for

Ethno Musicology

Problem Formulation

  • We investigate how deep learning can help identify features in music to identify the ethnic origin of the piece.
  • We attempt to analyze the music based on the geographical region it originated from as well as its time period. We try to identify influences from various regions in a piece of music.

Our current approach

  • We first train a deep neural network on data from the Smithsonian Global Sound database. We test with a subset of the data from the database that is not used for training.
  • The classifier network is convolutional and takes a spectrogram of the music audio as the input.