DESARROLLO DE UN ALGORITMO BASADO EN REDES NEURONALES PARA LA DETECCIÓN AUTOMÁTICA DEL TIPO DE INSTRUMENTO MUSICAL
Fecha
2018-04-23
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
[ES]Objetivo: Implementación de un sistema capaz de reconocer instrumentos musicales de forma automática, utilizando como herramienta las Redes Neuronales Artificiales.
La clasificación instrumental constará de varias partes. En primer lugar, llevaremos a cabo un procesado de señal para extraer parámetros característicos de notas musicales provenientes de violonchelo, viola, violín y flauta travesera. Extraeremos de manera numérica lo que para nosotros es el timbre de un instrumento, característica subjetiva del sonido que permite a nuestro cerebro saber diferenciar entre diferentes fuentes sonoras.
El timbre de un instrumento está relacionado con la disposición de los armónicos de cada nota musical y cómo varían sus amplitudes en el tiempo, se calcula pues la frecuencia fundamental de cada nota y a partir de ella se extrae la potencia relativa de dichos armónicos.
Será esto información valiosa para la red neuronal, que nos dirá qué instrumento suena tras haber sido entrenada de manera supervisada.
[EN]Objective: To implement a new system capable of recognising musical instruments automatically, using the Artificial Neuronal Networks as a tool. The instrumental classification will consist of several parts. Firstly, we will carry out a signal processing to extract characteristic parameters of the musical notes produced by a cello, a viola, a violin and a traverse flute. We will extract numerically the timbre of an instrument, a subjective feature of the sound, which allows our brain to distinguish between different sound sources. The timbre of an instrument is related to situation of the harmony in each musical note, and how their amplitude varies in time; thus, the fundamental frequency of each note is calculated and from it, the relative potential of these harmonies are extracted. This information will be valuable for our neuronal network, which will tell us which instrument is sounding after training it.
[EN]Objective: To implement a new system capable of recognising musical instruments automatically, using the Artificial Neuronal Networks as a tool. The instrumental classification will consist of several parts. Firstly, we will carry out a signal processing to extract characteristic parameters of the musical notes produced by a cello, a viola, a violin and a traverse flute. We will extract numerically the timbre of an instrument, a subjective feature of the sound, which allows our brain to distinguish between different sound sources. The timbre of an instrument is related to situation of the harmony in each musical note, and how their amplitude varies in time; thus, the fundamental frequency of each note is calculated and from it, the relative potential of these harmonies are extracted. This information will be valuable for our neuronal network, which will tell us which instrument is sounding after training it.
Descripción
Palabras clave
Sistemas de Telecomunación y Sonido e Imagen