Publicação

Signal classification based on a hybrid approach of supervised and unsupervised machine learning

Detalhes bibliográficos
Resumo:	The main challenge in using optical sensors to collect data for automated prosthesis control lies in accurately predicting which finger is moving based solely on forearm signals. This is a complex task in the design of intelligent prosthetic systems, which must be both efficient and adaptive. However, improving the quality of life for individuals with motor disabilities demands reliable interpretation of such biosignals. This work proposes the use of machine learning algorithms to address this problem. In this research was used a dataset of signal acquired with a Fiber Bragg Grating sensor positioned on the forearm, on a group of ten patients. The group were asked realize some finger movements in order to gather data. The main problem is to identify which movement is being realized without labeling the signal. In this research will be analyzed methods to apply label on the data and classify them. The focus was to get a precise hybrid approach of supervised and unsupervised methods. k-Means was used as an unsupervised machine learning method to group similar data into distinct clusters and label the data. Random Forest was used as supervised learning algorithms to classify the data after labeling.
Autores principais:	Vergilino, Gregory Marcelo
Assunto:	Clustering Biosignals k-Means Random forest
Ano:	2025
País:	Portugal
Tipo de documento:	dissertação de mestrado
Tipo de acesso:	acesso aberto
Instituição associada:	Instituto Politécnico de Bragança
Idioma:	inglês
Origem:	Biblioteca Digital do IPB

Descrição
Resumo:	The main challenge in using optical sensors to collect data for automated prosthesis control lies in accurately predicting which finger is moving based solely on forearm signals. This is a complex task in the design of intelligent prosthetic systems, which must be both efficient and adaptive. However, improving the quality of life for individuals with motor disabilities demands reliable interpretation of such biosignals. This work proposes the use of machine learning algorithms to address this problem. In this research was used a dataset of signal acquired with a Fiber Bragg Grating sensor positioned on the forearm, on a group of ten patients. The group were asked realize some finger movements in order to gather data. The main problem is to identify which movement is being realized without labeling the signal. In this research will be analyzed methods to apply label on the data and classify them. The focus was to get a precise hybrid approach of supervised and unsupervised methods. k-Means was used as an unsupervised machine learning method to group similar data into distinct clusters and label the data. Random Forest was used as supervised learning algorithms to classify the data after labeling.