Protótipo para auxílio de identificação do tipo de pisada baseada em sensores piezoelétricos e redes neurais artificiais
The aim of this work is the development of an instrumented insole, based on ceramic piezoelectric sensors and artificial neural networks to identify the type of footprint; and thus assist in the analysis and diagnosis of health specialists.The plantar pressure is the result of the contact of the pla...
| Autor principal: | Vieira, Mário Elias Marinho |
|---|---|
| Formato: | Dissertação |
| Idioma: | Português |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2019
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/3990 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
The aim of this work is the development of an instrumented insole, based on ceramic piezoelectric sensors and artificial neural networks to identify the type of footprint; and thus assist in the analysis and diagnosis of health specialists.The plantar pressure is the result of the contact of the plantar region (lower foot) with a surface, and can be measured by baropodometric platforms or instrumented insoles. This variable is used in studies of postural correction, movement analysis, correction of the type of footfall and identification of diseases in the plantar region. The device contains an instrumented insole with 13 sensors divided into plantar regions (hindfoot, midfoot, forefoot and hallux), coupled to a central board that performs conditioning and wireless transmission of the data. As a receiver and data storage, a mobile device (smartphone) was used via Bluetooth© communication. Plantar pressure data were collected from 14 people with mean age of 28.7 ± 8.8 years, used as input in an artificial neural network (RNA) MLP (Multi-Layer Perceptron) to classify the type of footprint of each one. All subjects performed a procedure of walking 10 times a course of 10 meters with the device installed, totaling 100 meters. After data collection, a health professional evaluated each individual to provide the desired output of RNA in supinated, pronated and neutral. The data were processed and divided into samples, which were used as RNA database. As methods of RNA training, holdout and cross validation were applied. The MLP containing 21 neurons in the hidden layer, using cross-validation, obtained a 99.63% accuracy for the entire data set. In addition, the obtained results are divided in accuracy, sensitivity, specificity and efficiency (confusion matrix) by type of footprint; showing the lower and higher values respectively: 99.7% e 100% in accuracy (neutral, supinated and pronated), 99.7% and 100% in sensitivity (neutral and pronated), 99.8% and 100% in specificity (supinated and pronated), and finally, 99.7% and 100% in efficiency (supinated and pronated). Based on these values, it can be stated that RNA presented the best performance in the classification of the type of footprint in pronated and there is a greater difficulty to distinguish supinated from neutral. It is concluded that it is possible to use the device developed for plantar pressure measurement, as well as the MLP designed to classify the type of footprint. |
|---|