eduzhai > Applied Sciences > Engineering >

wav2shape Hearing the Shape of a Drum Machine

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: Disentangling and recovering physical attributes, such as shape and material,from a few waveform examples is a challenging inverse problem in audio signalprocessing, with numerous applications in musical acoustics as well asstructural engineering. We propose to address this problem via a combination oftime--frequency analysis and supervised machine learning. We start bysynthesizing a dataset of sounds using the functional transformation method.Then, we represent each percussive sound in terms of its time-invariantscattering transform coefficients and formulate the parametric estimation ofthe resonator as multidimensional regression with a deep convolutional neuralnetwork. We interpolate scattering coefficients over the surface of the drum asa surrogate for potentially missing data, and study the response of the neuralnetwork to interpolated samples. Lastly, we resynthesize drum sounds fromscattering coefficients, therefore paving the way towards a deep generativemodel of drum sounds whose latent variables are physically interpretable.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...