Please use this identifier to cite or link to this item:
http://hdl.handle.net/10609/119986
Title: | Sistema de detección de instrumentos musicales en señales de audio |
Author: | Juan Monguillot, Marc |
Tutor: | Vilaplana Besler, Veronica |
Others: | García-Solórzano, David |
Abstract: | This project presents the design of a musical instrument detection system developed with the innovative requirement of having some ability to distinguish between different categories mixed simultaneously in the same signal, without restrictions regarding possible combinations. The approach followed is based on the fact that the spectrum of musical notes is constrained to the regions near the fundamental frequency and its harmonics. This allows, under certain conditions, to separate the approximate spectrum of different notes mixed in a signal excerpt. For this purpose, a set of features calculated from the separated spectrum of each note is proposed so that several vectors can be extracted from the same signal excerpt. An algorithm has also been developed to identify the different overlapping notes by precisely detecting their fundamental frequencies. To classify the vectors, a support vector machine (SVM) with cubic Kernel has been chosen after carrying out various comparison and optimization tests with different machine learning techniques. Finally, a MATLAB App has been developed as a prototype to test the system. This model has been trained for 26 different instruments, obtaining approximate accuracy rates of 90% for signals with a single instrument and around 60% for signals containing between 2 and 4 instruments. |
Keywords: | digital audio processing machine learning pattern recognition |
Document type: | info:eu-repo/semantics/bachelorThesis |
Issue Date: | Jun-2020 |
Publication license: | http://creativecommons.org/licenses/by-nc-sa/3.0/es/ |
Appears in Collections: | Bachelor thesis, research projects, etc. |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Presentación.mp4 | Presentación del TFG | 118,61 MB | MP4 | View/Open |
Clasificador de instrumentos.mlappinstall | Instalador de la MATLAB App con el prototipo del sistema entrenado para 26 instrumentos | 279,16 MB | MATLAB App Installer | View/Open |
Código MATLAB.zip | Archivos de código MATLAB del sistema diseñado | 286,72 kB | Carpeta comprimida (en ZIP) | View/Open |
mjuanmonTFG0620memòria.pdf | Memoria del TFG | 2,44 MB | Adobe PDF | View/Open |
Share:
This item is licensed under a Creative Commons License