Breathing and sound monitoring
Data
2022-12-09
Autores
Título da revista
ISSN da revista
Título do Volume
Editora
Resumo
Atualmente, há um número crescente de pessoas a viver com doenças mentais. Algumas das mais prevalentes são os distúrbios de ansiedade, estimando-se, em 2017, que cerca de 284 milhões de pessoas viviam algum tipo destes distúrbios. Esta dissertação sensibiliza para o tema desta doença de saúde mental e aborda o desafio da deteção de stress/ansiedade através de um sistema de classificação de ansiedade utilizando o método de machine learning supervisionado. É dado um especial foco ao sistema respiratório e aos seus parâmetros que se correlacionam com o stress/ansiedade. O trabalho desenvolvido estabelece uma framework para o desenvolvimento de um sistema de monitorização de ansiedade usando múltiplos parâmetros fisiológicos. Cinco dos algoritmos mais comuns para este tipo de trabalhos foram usados para a tarefa de classificação e aquele que alcançou os melhores resultados foi o algoritmo Random Forest com 92% acurácia e ótimos valores de precisão, “recall”, “f1-score” e “cohen kappa score”. Em última análise, esta tecnologia pode ser aplicada para fins de deteção de stresse/ansiedade de maneira autónoma e em parceria com especialistas que lidam com estes problemas diariamente, como psicólogos ou psiquiatras.
Nowadays, there is an increasing number of people living with mental illnesses. Some of the most prevalent are the anxiety disorders, being estimated that, in 2017, around 284 million people were living with these disorders. This dissertation raises awareness for this mental illness and addresses the challenge of stress/anxiety detection using a supervised machine learning system for classification. It’s given a special focus on the respiratory system and its parameters that correlate with stress/anxiety. The developed work establishes the framework for an anxiety monitoring system using multiple physiological parameters. Five of the most common algorithms were used for the task and the one achieving the best results was the Random Forest classifier with 92% “accuracy” and great values for “precision”, “recall”, “f1-score” and “cohen kappa score”. Ultimately, this technology can be applied to self and autonomous stress/anxiety detection purposes partnering it with specialists who deal with these problems daily, like psychologists or psychiatrists.
Nowadays, there is an increasing number of people living with mental illnesses. Some of the most prevalent are the anxiety disorders, being estimated that, in 2017, around 284 million people were living with these disorders. This dissertation raises awareness for this mental illness and addresses the challenge of stress/anxiety detection using a supervised machine learning system for classification. It’s given a special focus on the respiratory system and its parameters that correlate with stress/anxiety. The developed work establishes the framework for an anxiety monitoring system using multiple physiological parameters. Five of the most common algorithms were used for the task and the one achieving the best results was the Random Forest classifier with 92% “accuracy” and great values for “precision”, “recall”, “f1-score” and “cohen kappa score”. Ultimately, this technology can be applied to self and autonomous stress/anxiety detection purposes partnering it with specialists who deal with these problems daily, like psychologists or psychiatrists.
Descrição
MASTER in Biomedical Engineering, in accordance with the provisions of the General Regulation of Cycles of Studies leading to a Master's degree at UTAD DR, 2nd series No. 133 Regulation No. 658/2016 of July 13, 2016
Palavras-chave
Doenças de saúde mental , Stress