TRIPLE F (F³) FILTERING OF KINEMATIC DATA

Abstract

Raw data in biomechanical studies usually require filtering. Depending on the used filter there exist some drawbacks as “signal shifted relative to the raw data”, “instability or degradation”, “endpoint problem”, “oscillations”. A filter called triple F or F³ based on the Fourier transformation is presented that is not crippled by these drawbacks. It consist of a transformation of the original data into the frequency spectrum followed by eliminating the unwanted frequencies (windowing) and an inverse Fourier transformation back to the data as a function of time. This procedure is stable and does not shift the data. It is shown how to dampen additional oscillation on the filtered data and how to avoid the endpoint problem completely. A comparison with a Butterworth filter and an application completes the presentation.