Classificação de distúrbios vocais utilizando redes neurais artificiais

Abstract

The voice is an important means of communication, affecting the performance of professionals in all areas, requiring extreme care with vocal health. However, there are many challenges in voice assessment methods, providing research opportunities for non-invasive approaches. In line with this requirement, this paper investigates the application of artificial neural network to classify patterns related to voice disorders in speech signals. It is proposed to use the wavelet packet transform to extract signal characteristics in time-frequency analysis, artificial neural networks for recognition and classification of the patterns presented by vocal disorders. Finally, the modeling intelligent classifier has been shown to be consistent in the developed tests, proving itself promising for a future application as a non-invasive method that assists voice professionals.