Separación de fuentes sonoras en señales acústicas

Abstract

It can be considered as a sound source each element which generates any sound that can be detecter by a microphone. Here, the main objective is to develop separation techniques and to assess which kind of additional information, apart from the input signal, could help to obtain better results. This thesis is based on the hypothesis that the use of spectral instrument models, which have been proven as useful information at other signal processing tasks, could be a good tool to discriminate what instrument is the owner of a concrete piece of energy from the input signal. Then better results than the actual ones could be obtained. Besides, other basic hypothesis of this thesis is that a higher isolation ofthe notes of each instrument could be achieved ifthe overlapped harmonic partial problem were solved.