Sporathe knowledge network
ES
Universidad de San Buenaventura

Ingeniería de Sonido · 2017

Identificación automática de eventos sonoros de música y tráfico en grabaciones en ambientes controlados mediante descriptores de audio y el algoritmo de clasificación de máquinas de vectores de soporte (SVM)

Audio content analysis of real world recordings instead of common SPL measures is proposed as a technique for environmental noise evaluation. However, because of the large quantity of data that a monitoring system produces an automatic sound event algorithm was built using audio descriptors as MFCC, spectral centroid, zcr and roll off coupled with a supervised learning approach with SVM. For traffic related audio events an 84% recognition rate was achieved and for music related events a 95% rate in controlled environments. This was done in database composed of outdoor residential audios recorded in Medellin city.

Full text 78 pages with text

Read the full thesis Record in the repository

Contents

Keywords

Its library's thesaurus