Avaliação de sinais acústicos para o pré-diagnóstico de covid-19

Carregando...
Imagem de Miniatura

Data

2022-01-06

Autores

Nunes, Gabriel Andrey Perego

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Estadual Paulista (Unesp)

Resumo

The signals produced by the human body have been the subject of researches on indicators related to different diseases. With the outbreak of the Sars-Cov-2 pandemic in 2020, there is a unique opportunity for signal processing to contribute to the identification of the pathology. The first papers on the new virus were based on recent studies of other diseases that have similar symptoms to those caused by the coronavirus. Since the disease has common symptoms in patients and these symptoms can cause some degree of impairment in the vocal system and generate irregular vibrations, this study aims to identify Covid-19 signatures in audios with sustained vowels through analysis of the energy in the sound waves. By using Wavelet transform and Teager operator, it was possible to identify, with statistical methods, unique characteristics of healthy and infected individuals, which can be used in the creation of classifiers to identify the pathology. It was found that infected people seems to have a vector with an average value of 0.00034 while healthy people produce a value of 0.01665. Also, it was possible to analyze the difference between the minimum and maximum values found in the vectors.
Os sinais produzidos pelo corpo humano têm sido objeto de pesquisa na busca de indicadores relacionados às mais diversas doenças. Com a explosão da pandemia do vírus Sars-Cov-2 em 2020, surge a oportunidade ímpar do processamento de sinais contribuir na identificação da patologia. Os primeiros trabalhos relacionados ao novo vírus se basearam em estudos recentes de outras doenças que dispõem de sintomas similares aos causados pelo Coronavírus. Dado que a doença apresenta sintomas comuns em pacientes, e que esses sintomas podem causar algum grau de comprometimento no sistema fonador e gerar vibrações irregulares, o presente trabalho teve por objetivo identificar assinaturas da covid-19 em áudios de vogais sustentadas por meio da análise da energia presente nas ondas sonoras. Utilizando Transformada de Wavelets e o operador de Teager, foi possível, com métodos estatísticos, identificar características únicas de indivíduos saudáveis e de infectados que podem ser utilizadas na criação de classificadores para identificação da patologia. Foi descoberto que pessoas infectadas apresentam um valor médio no vetor resultante de 0,00034 enquanto pessoas saudáveis obtiveram um valor de 0,01665. Foi possível também analisar a diferença entre os valores mínimos e máximos encontrados nos vetores.

Descrição

Palavras-chave

Signal processing, Wavelet transform, Pathology identification, Teager operator, Processamento de sinais - Técnicas digitais, Transformada de Wavelets, Identificação de patologia, Operador de Teager

Como citar