Análise de sinais pulmonares utilizando técnicas no domí­nio Tempo-Frequência e Classificação Neural

• Ano de defesa: 2010
• Autor(a) principal: Almeida, Alberto Jorge Santos de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Alagoas; BR; Programa de Pós-Graduação em Modelagem Computacional de Conhecimento; UFAL
• Palavras-chave em Português: Sings Lung; Wavelet transform; Neural Network; Sinais Pulmonares; Transformada Wavelet; Redes Neurais; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufal.br/handle/riufal/834

Auscultation is a method of clinical practice, simple, noninvasive, used to diagnose diseases of the respiratory system. However, it is an imprecise method because, among other factors, the limitations of the auditory system, the overlap of heart sounds and human hearing sensitivity difference, besides the limited spectral response characteristic of many commercial stethoscopes. These factors contribute to the diagnosis relies heavily on the experience of the professional expert. The acoustic analysis of spectral characteristics of signals of ventilation can be a complementary diagnostic technique in facilitating the process of detection and identification of breath sounds, providing aid in the assessment of symptom severity and treatment efficacy. In this study, we attempted to structure a process of analysis of pulmonary signs to identify characteristics of respiratory disorders. Accordingly, the signals were processed by filtering processes and decomposed into sub-frequency bands through discrete wavelet transform (DWT), generating vectors as coefficients for classification using an Artificial Neural Network. A case study with signals obtained from tests was presented and duly considered.