Music Structure Boundaries Estimation Using Multiple Self-Similarity Matrices as Input Depth of Convolutional Neural Networks - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2017

Music Structure Boundaries Estimation Using Multiple Self-Similarity Matrices as Input Depth of Convolutional Neural Networks

Alice Cohen-Hadria
  • Fonction : Auteur
  • PersonId : 999469

Résumé

In this paper, we propose a new representation as input of a Convolutional Neural Network with the goal of estimating music structure boundaries. For this task, previous works used a network performing the late-fusion of a Mel-scaled log-magnitude spectrogram and a self-similarity-lag-matrix. We propose here to use the square-sub-matrices centered on the main diagonals of several self-similarity-matrices, each one representing a different audio descriptors. We propose to combine them using the depth of the input layer. We show that this representation improves the results over the use of the self-similarity-lag-matrix. We also show that using the depth of the input layer provide a convenient way for early fusion of audio representations.
Fichier principal
Vignette du fichier
music-structure-boundaries.pdf (334.97 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01534850 , version 1 (08-06-2017)

Identifiants

  • HAL Id : hal-01534850 , version 1

Citer

Alice Cohen-Hadria, Geoffroy Peeters. Music Structure Boundaries Estimation Using Multiple Self-Similarity Matrices as Input Depth of Convolutional Neural Networks. AES International Conference Semantic Audio 2017, Jun 2017, Erlangen, Germany. ⟨hal-01534850⟩
181 Consultations
852 Téléchargements

Partager

Gmail Facebook X LinkedIn More