Comparative evaluation of techniques for the harmonic analysis of human motion data.

Four mathematical techniques for the estimation of the Fourier coefficients of pseudoperiodic non-exact discrete functions were submitted to comparative evaluation. These techniques were devised within the following basic procedures: (1) numerical harmonic analysis applied either directly or after redistribution of the data points through some interpolation procedure, so that they be evenly spaced in time and exactly fit on cycle period; (2) fitting of the empirical data with an analytical model followed by the calculation of the Fourier integrals of this model. The evaluation was carried out with special reference to the use of these techniques for processing human motion photogrammetric data. The following criteria were used: (1) accuracy of the Fourier coefficient estimates, (2) capability of yielding information about this accuracy, (3) a priori information needed regarding the statistical properties of the experimental error, (4) factors concerning implementation in digital computers. Practical information was obtained for the non-specialist user with regard to the choice of one technique among the several possible in particular circumstances.