Comparaison de Divers Algorithmes : Recherche sur l'Identification des Caractéristiques des Signaux Audio de Piano

Auteurs-es

Mots-clés :

piano, caractéristique audio, réseau de neuronal conventionnel, déformation temporelle dynamique

Résumé

This article briefly introduced feature extraction methods for piano audio signals and algorithms based on dynamic time warping (DTW), back-propagation neural network (BPNN), and convolutional neural network (CNN), which can recognize piano audio features. The three recognition algorithms were compared in the subsequent simulation experiments. It was found that for some single-note and multi-note piano audios, the recognition results of the CNN algorithm were consistent with the standard results, the BPNN algorithm had some differences, and the DTW-based recognition algorithm had the most differences. As the number of notes in the piano audio increased, the recognition accuracy of all the algorithms decreased, but the CNN algorithm decreased the least, and its recognition performance was highest under the same number of notes, followed by the BPNN algorithm, and the DTW-based recognition algorithm was the lowest.

Fichiers supplémentaires

Publié-e

2023-08-24

Comment citer

1.
Hao S. Comparaison de Divers Algorithmes : Recherche sur l’Identification des Caractéristiques des Signaux Audio de Piano. Canadian Acoustics [Internet]. 24 août 2023 [cité 6 oct. 2024];51(2). Disponible à: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/3941

Numéro

Rubrique

Article - Traitement des signaux / Méthodes numériques