Metodologia para a suavização de dados biomecanicos por função não parametrica ponderada local robusta

AUTOR(ES)
DATA DE PUBLICAÇÃO

1998

RESUMO

Human movement is a complex phenomenon that must be studied by many areas. Biomechanics works together with other areas to search a better insight of the human movement. Many times, the kinematics analysis produce discrete data and -, one needs to adjust a continuous function to represent them with accuracy. If the data are in time series, and the curves family that is formed has a probability distribution, it is called stochastic processoTo smooth the time series data that have a stochastic dependence between the nearest values, means to leave the data show their adjustment curve shape. The non parametric procedures, partial adjustments for instance, allow that the smoothed data to receive a big influence from the neighbors and none influence from the apart points. There are a lot of functions that can be found in the literature to smooth discrete data. The aim of this work is to show a methodology for the robust local weighted non parametric function to smooth biomechanics discrete data to describe continuous human movements. The methodology consists in choosing the window size (smoothing parameter f) in witch the weighted regression will be held, the weight function that will be used (in this case, the tricubic weight), the polynomial degree adjusted in that region (second degree for this work) and compare the residuais values with the biquadratic estimation for a possible reweight in the regression to eliminate the outliers. This window moves itself so that it can smooth each point of the data vector, which ensemble will be interpolated by a continuous function that has a continuous second derivative. Near the edge points the window assumes an asymmetric form and there is a participation of the remote data, but the tricubic weight solves this problem. We show four examples of application of loess function to smooth biomechanics data: an angular movement of a boy s left shank during a cycle of a race, the horizontal movement of the right ankle of a subject during a soccer kick session with stopped ball, the goalkeeper horizontal displacement during a specific training test of defenses, and the professional soccer player movement during an official game. In ali of these cases, the researcher chose the smooth parameter (f) that provide an adequate smoothing, that were proved by the residuais analysis, by the data autocorrelation analysis and by the derivatives behaviors analysis. The loess function revealed itself very appropriate in kinematics data smoothing of the human movement, mainly depict of the robust procedure to identify outliers and to correct data weight

ASSUNTO(S)

cinematica movimento biomecanica

ACESSO AO ARTIGO

Documentos Relacionados