Visualization of audio data using stacked graphs

Mathieu Lagrange (1) , Mathias Rossignol , Grégoire Lafay (1)

1 LS2N - Laboratoire des Sciences du Numérique de Nantes

In this paper, we study the benefit of considering stacked graphs to display audio data. Thanks to a careful use of layering of the spectral information, the resulting display is both concise and intuitive. Compared to the spectrogram display, it allows the reader to focus more on the temporal aspect of the time/frequency decomposition while keeping an abstract view of the spectral information. The use of such a display is validated using two perceptual experiments that demonstrate the potential of the approach. The first considers the proposed display to perform an identification task of the musical instrument and the second considers the proposed display to evaluate the technical level of a musical performer. Both experiments show the potential of the display and potential applications scenarios in musical training are discussed.

Domaines: Machine Learning [stat.ML]; Traitement du signal et de l'image [eess.SP]; Intelligence artificielle [cs.AI]; Apprentissage [cs.LG]; Multimédia [cs.MM]; Traitement du signal et de l'image [eess.SP]

Mathieu Lagrange, Mathias Rossignol, Grégoire Lafay. Visualization of audio data using stacked graphs. 19th International Society for Music Information Retrieval Conference, Sep 2018, Paris, France. ⟨hal-01807481⟩