bachelorThesis
Classificação automática de instrumentos musicais utilizando uma abordagem de classificação multivariada de séries temporais
Fecha
2022-06-27Registro en:
BABINSKI, Welliton Jhonathan Leal. Classificação automática de instrumentos musicais utilizando uma abordagem de classificação multivariada de séries temporais. 2022. Trabalho de Conclusão de Curso (Bacharelado em Engenharia de Computação) - Universidade Tecnológica Federal do Paraná, Pato Branco, 2022.
Autor
Babinski, Welliton Jhonathan Leal
Resumen
This work presents a development of an essential algorithm to automatic audio mixing and repair intelligent systems, is a machine learning model that classifies automatically musical instruments using a multivariate time series classification approach. Starting from two databases of audio digital signals with various musical instruments samples, the first goal is to extract both temporal and spectral instantaneous features, which are represented by temporal series. The series are used to train the supervised machine learning models, that are responsable to identify patterns of temporal series using adapted algorithms to do this task, like the K-Nearest Neightbours with the Dynamic Time Warping algorithm, or the Support Vector Machines with Global Alignment Kernel algorithm. The main goal is to utilize this models to do the task of feature classification of new databases of unknow musical instruments audio signal, as well to analysis and understand which effects different signal durations and feature extraction parameters have in the results.