MusicFactoryApplication of a Convolutional Neural Network for the Generation of Soundscapes from Images

  1. Juan José Navarro Cáceres
  2. André Filipe Sales Mendes
  3. Hector Sánchez San Blas
  4. Gabriel Villarrubia González
  5. María Navarro Cáceres
Book:
New Trends in Disruptive Technologies, Tech Ethics and Artificial Intelligence: The DITTET 2022 Collection
  1. Daniel Hernández de la Iglesia (ed. lit.)
  2. Juan Francisco de Paz Santana (ed. lit.)
  3. Alfonso José López Rivero (ed. lit.)

Publisher: Springer International Publishing AG

ISBN: 978-3-031-14858-3

Year of publication: 2022

Pages: 156-164

Congress: DiTTEt: International Conference on Disruptive Technologies, Tech Ethics and Artificial Intelligence (2. 2022. Salamanca)

Type: Conference paper

Abstract

A soundscape is a sound description of a concrete environment. Therefore, the soundscapes are always connected to a visual component, as it might capture sounds from an urban city, a countryside, or a domestic place. In this work, we present a system that generate soundscapes from images. Firstly, we recognize some objects in the image. In a second step the system searches the most adequate sounds according to the entities identified in the picture. Finally, a soundscape is synthesized by combining the short sound files found. The results obtained according to the subjective evaluation are promising and encouraging to deepen our research in the soundscape generation.