Classificação automática de desordens vocais usando a variância wavelet

• Ano de defesa: 2022
• Autor(a) principal: Santos, Rafael Alberto dos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Cornelio Procopio; Brasil; Programa de Pós-Graduação em Engenharia Elétrica; UTFPR
• Palavras-chave em Português: Wavelets (Matemática); Reconhecimento automático da voz; Classificação; Wavelets (Mathematics); Automatic speech recognition; Classification; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/30196

Vocal disorders may be present when the voice fails to fulfill its basic role of verbal and emotional transmission. These disturbances can be perceived by the variation of perceptual parameters of the voice, such as quality, pitch, and loudness. Changes in voice parameters can be measured and classified automatically through acoustic analysis. The present work proposes an algorithm for automatic classification of voice disorders, using wavelet variance in signals of vowel "a" with neutral pitch to form a feature vector. The pathology under analysis is nodules and Reinke's edema. These pathologies affect the vocal folds and alter acoustic parameters of voice signals. Classification is performed using a supervised learning technique called support vector machine. The experiments are performed as a binary classification between the groups Edema/Healthy, Nodule/Healthy, Edema/Nodule and Pathological/Healthy, being the pathological class formed by the pathologies nodule and Reinke's edema. In order to compare the results, the extraction of features of the voice signals is carried out with two other methods, the mel spectrogram and the mel frequency cepstral coefficients. The results obtained in the tests are promising and indicate that the features extracted from the signals using wavelet variance discriminate the classes and can replace the mel spectrogram and MFCC techniques.