We present SAMPLE, a Python package of tools for spectral analysis and modal parameters estimate. The core of the package is an implementation of the “Spectral Analysis for Modal Parameters Linear Estimate” (SAMPLE) algorithm. This includes a custom implementation of a Sinusoidal Analysis algorithm based on Spectral Modelling Synthesis. Our custom implementation is specifically designed for modal tracking. We also included utilities for automatically tuning the algorithm parameters, using a Bayesian optimization method based on Gaussian Processes. For this purpose, we implemented efficient routines for computing perceptual audio representations for loss functions, such as the multiscale-spectrogram, the mel-spectrogram and the cochleagram. The package also comes with a Graphical User Interface, which allows to load and trim audio inputs, set the algorithm parameters, run the algorithm, listen to a resynthesis of the input, and export the results. The GUI is distributed both as an extra for the Python package and as a standalone executable.
SAMPLE: a Python Package for the Spectral Analysis of Modal Sounds / M. Tiraboschi, F. Avanzini - In: Corpi fisici = Physical bodies[s.l] : AIMI, 2022. - ISBN 9788890341366. - pp. 50-55 (( Intervento presentato al 23. convegno Colloquium on Music Informatics tenutosi a Ancona nel 2022.
SAMPLE: a Python Package for the Spectral Analysis of Modal Sounds
M. Tiraboschi
Primo
;F. AvanziniUltimo
2022
Abstract
We present SAMPLE, a Python package of tools for spectral analysis and modal parameters estimate. The core of the package is an implementation of the “Spectral Analysis for Modal Parameters Linear Estimate” (SAMPLE) algorithm. This includes a custom implementation of a Sinusoidal Analysis algorithm based on Spectral Modelling Synthesis. Our custom implementation is specifically designed for modal tracking. We also included utilities for automatically tuning the algorithm parameters, using a Bayesian optimization method based on Gaussian Processes. For this purpose, we implemented efficient routines for computing perceptual audio representations for loss functions, such as the multiscale-spectrogram, the mel-spectrogram and the cochleagram. The package also comes with a Graphical User Interface, which allows to load and trim audio inputs, set the algorithm parameters, run the algorithm, listen to a resynthesis of the input, and export the results. The GUI is distributed both as an extra for the Python package and as a standalone executable.File | Dimensione | Formato | |
---|---|---|---|
XXIII_CIM_2022_paper_5979.pdf
accesso aperto
Tipologia:
Publisher's version/PDF
Dimensione
398.46 kB
Formato
Adobe PDF
|
398.46 kB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.