REVISIÓN ANALÍTICA DE 3 ALGORITMOS DE RECONOCIMIENTO DE CONTEXTO DE RUIDO PARA LA DEFINICIÓN DE UNA PROPUESTA DE UN ALGORITMO EMPÍRICO.

Abstract

The main objective of the present degree work is to propose a context recognition algorithm for background noises through a comparative analysis of algorithms developed by other researchers, where non-speech sound recognition techniques will be reflected from a point of view. practical and functional sight. The methodology used for the development of this research is exploratory, qualitative and applied, for which it was necessary to apply a strategy based on the collection of information necessary for the development of an algorithm. Four background noise reduction techniques were reviewed: the adaptive RACE (correlation adaptive enhancement) filter, spectral subtraction, GSVD (generalized singular value decomposition) filtering, and generalized singular value decomposition. of independent components. The first thing that was done was to decompose the audio signal into a single channel, then through functions, the noise of the signal was detected by using the fast Fourier transform, the filters for DSP (digital signal processing) and finally the separation of the background noise was achieved.