Comparação de dois métodos para representação da força de reação do solo no desempenho de classificação de padrões da marcha

Abstract

The influence of different input data (Fourier versus Legendrecoefficients), on gait patterns classification performance were analyzed byOrdinary Least Squared Method (OLS) and Least Absolute Shrinkage andSelector Operator (LASSO). The classification performance was tested inbarefoot gait and shod gait of forty-three individuals, 19 men and 24 women, 24,8 years old, walking on a force platform at a self-determined velocity and 1,3 m.s-1. Every individual performed 40 trials barefoot and 40 trials with their own shoe wear. The vertical component of the ground reaction force was registered at frequency of 1 kHz. The Legendre coefficients and the Fourier coefficients were applied as a standard entry for the OLS and the LASSO. The OLS analysis of vertical component of ground reaction force (Fy), suggested the recognition rates (e.g. barefoot gait versus shod gait) were higher than 99,0% for both types of input data. The recognition rate with the Lasso, for vertical component of ground reaction force (Fy), was higher than 98,0% using Fourier coefficients and higher than 99% for Legendre coefficients. The coefficient number used by the LASSO is smaller than OLS coefficient number.