Aprendizagem profunda aplicada na classificação de Taikos

Abstract

Deep learning is a part of machine learning, it has the function of enabling a machine from received data, the resolution of designated problems without the need of constant supervision and human intervention. The differential of deep learning is its ability to divide a complex problem into several layers that subdivided in a simple way, facilitate the analysis, classification and solution. Taking this aspect of deep learning into account, the proposal for this assignment is to apply it in the recognition of percussion musical instruments, the type and intensity of percussion being applied, from the sound waves emitted by them, in a simplified way, audio tracks of the instruments will be recorded, for system training, the algorithms after training will be able to identify the particularities of each instrument and beat and thus classify them. The instruments to be used are the variants of taikos, the Japanese drums, at first the types considered are the shime daiko, okedo daiko, hira-daiko and nagadodaiko. The aim is the construction of the neural network and the treatment of the signal using the python programming language.