Classification of Quiet Standing Control by Perceptron Neural Network and Principal Component Analysis: a Pilot Study

This work aims at classifying the body sways based on principal components of elliptical sway area (EA) and mean velocity (MV) of center of pressure by perceptron neural network. A sample of 27 young, healthy male adults was monitored during a stabilometric test, standing on a force platform during 3 min, with eyes closed, and feet in a closed position. The data were stored in a matrix, where rows represent subjects and the columns the sway path, MV, both in mediolateral and anterior-posterior axis, and EA. Then, PCA was applied. The median of MV and EA were used for performing the separation in two groups with lower and greater values than median. The perceptron neural network was performed to classify subjects based on sort values of MV and EA. The overall explained variance was 90% with 76% in the first component. The perceptron neural network classified linearly the subjects in two groups, with greater error when the EA was the discriminant variable than MV. The mean velocity was more effective as a target during the performance of the neural network.