Classification of Voice Pathologies Using Glottal Signal Parameters

Manoela Kohler, Leonardo Alfredo Forero Mendoza, Juan G. Lazo , Marley Vellasco, Edson Cataldo

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

The classification of voice diseases has many applications in health, disease treatment, and the projection of new medical equipments for diagnosing these pathologies. This work uses the parameters of the glottal signal that are more likely to identify two types of voice disorders: vocal cord nodule and unilateral paralysis of vocal cords. The parameters of the glottal signal are obtained through the known inverse filtering method. The parameters of the glottal signal serve as input to a neural network that classifies into three different groups of speakers: speakers with pathology nodule on one’s vocal cords; with
unilateral vocal cord paralysis; and finally speakers with normal voices. The database is composed of 248 voice recordings containing samples of the three groups mentioned. In this study we have used a larger database for the classification compared with similar studies, and its classification rate is superior to other studies, reaching 95.83%.
Original languageEnglish
Title of host publicationAnais do 10. Congresso Brasileiro de Inteligência Computacional
Number of pages8
DOIs
StatePublished - 2011
Event10th Brazilian Congress on Computational Intelligence (CBIC’2011) - Ceará, Brazil
Duration: 8 Nov 201111 Nov 2011

Congress

Congress10th Brazilian Congress on Computational Intelligence (CBIC’2011)
Country/TerritoryBrazil
CityCeará
Period8/11/1111/11/11

Cite this