Classification of voice aging using parameters extracted from the glottal signal

Authors:
Leonardo Alfredo Forero Mendoza;Edson Cataldo;Marley Vellasco;Marco Silva
Affiliations:
Eletrical Engineering Department, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brasil;Applied Mathematics Department-Telecommunications Engineering, Universidade Federal Fluminense, Niterói, RJ, Brasil;Eletrical Engineering Department, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brasil;Eletrical Engineering Department, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brasil
Venue:
ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part III
Year:
2010

Citing 2
Cited 1

Glottal wave analysis with Pitch Synchronous Iterative Adaptive Inverse Filtering

Speech Communication - Eurospeech '91
Amplitude domain quotient for characterization of the glottal volume velocity waveform estimated by inverse filtering

Speech Communication

Sound recognition

Computing with instinct

Quantified Score

Hi-index	0.00

Visualization

Abstract

Classification of voice aging has many applications in health and geriatrics. This work focuses on finding the most significant parameters to identify voice aging. This work proposes to choose the most significant parameters extracted of the glottal signal to identify the voice aging process of men and women using the wrapper approach combining a genetic algorithm (as a search algorithm) with a neural network (as an induction algorithm). The chosen parameters will be used as entries in a neural network to classify male and female Brazilian speakers in three different age groups, which will be called young (from 15 to 30 years old), adult (from 31 to 60 years old) and senior (from 61 to 90 years old). The voice database used for this work was composed by one hundred twenty Brazilian people (male and female) of different ages. In this work we use the largest basis for classification of age compared with other similar works, and its rate of classification is superior to other studies reaching 91.6% in males and 83.33% in women.