Luva instrumentada para reconhecimento de padrões de gestos em Libras
This paper presents the development of a system to recognize gestures patterns of the Brazilian Sign Language (Libras). This system is composed of an instrumented glove, acquisition system, processing and classification by Artificial Neural Networks (RNA). The developed glove has five flex-sensors,...
| Autor principal: | Dias, Thiago Simões |
|---|---|
| Formato: | Dissertação |
| Idioma: | Português |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2020
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/5018 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
This paper presents the development of a system to recognize gestures patterns of the Brazilian Sign Language (Libras). This system is composed of an instrumented glove, acquisition system, processing and classification by Artificial Neural Networks (RNA). The developed glove has five flex-sensors, two contact sensors and an inertial sensor (three-axis accelerometer and gyroscope). Two versions of data acquisition systems were used to collect the data regarding the gestures performed by volunteers: wired data acquisition system and wireless data acquisition system. In the wired system, five volunteers participated in the collection of data related to the characters of Libras alphabet. With the wireless system, ten volunteers participated in the collection of ten different words in Libras. The collected data were segmented in three windows (fixed amounts of signal samples) that represent the construction period, gesture period and relaxation period of the gestures. After the segmentation, each segmented window was submitted to the extraction of features to generate a vector of features. For classification, the vector of features was divided into 80% for training and 20% for testing of the RNA. The accuracy rate obtained for manual alphabet gestures was 96.19% and the accuracy rate obtained for word gestures was 98.96%. During the research processes, some contributions were generated through the performed analysis, evidencing the potential of the system to gestures recognize in Libras. The performed and discussed analyses in the work are related to sensors, characteristics, volunteers and amount of separate data for network training. |
|---|