Software inteligente para el procesamiento musical de guitarra melódica aplicando una red neuronal supervisada.

In the last years, the computer has pointed to develop Artificial Intelligence (AI) as it was enunciated in its beginnings, machines that can think like humans. A wide field of application is the music, because the hardware and software tools facilitate the capture and audio analysis. The purpose of this work is to help the melodic guitar musical composition through an application that captures audio from the played melody, and, after making the respective frequency analysis through a supervised neural network, gives the musician an answer about the tone of the piece and the possible structure to follow, so this piece of music is pleasing to the ear, from the point of view of Western music. In the software development XP methodology was applied, adapting it to the characteristics of the project, specifically the development of the neural network, where the flow of work proposed by the creators of the Neural Network Toolbox of Matlab was followed. The application was coded using Matlab, because it’s offered toolboxes for signal processing and for the creation of neural networks, facilitate the implementation of transfer functions and algorithms for training and learning of the network. The architecture of the neural network is a Multilayer Perceptron and the dataset for training was developed with 180 pieces of classical music

• Hovedforfatter: Granda Veintimilla, Rodrigo Cornelio, Juan Guillermo Pinzón González (author)
• Format: bachelorThesis
• Sprog: spa
• Udgivet: 2015
• Fag: <DSP> <RED NEURONAL> <MULTILAYER PERCEPTRON> <TRANSFORMADA DE FOURIER> <INTELIGENCIA ARTIFICIAL> <ESCALAS MUSICALES> <CAPTURA DE AUDIO> <MATLAB> <GUITARRA> <BACK PROPAGATION>
• Online adgang: http://dspace.unl.edu.ec/jspui/handle/123456789/11721