Smoothing of GRF data using functional data analysis technique

Filtering is necessary in most digital signal processing. Data generated by mechanical equipment usually contains noise or spikes that need to be filtered before further processing occurs. This paper describes a method for smoothing or filtering spikes or noise present in the data by using a functional data analysis approach applied to biomechanical ground reaction force data. Ground reaction force data were collected from ten military subjects, age 31 ± 6.2 years, weight 71.6 ± 10.4 kg and height 166.3 ± 5.9 cm, using the Vicon 1.4 motion analysis system, Kistler force plates and thirty nine body markers. The results show that the optimum smoothing for this kind of data is obtained using a B-spline basis, penalizing fourth derivatives and a smoothing amount lambda equal to 1e-12. Functional data analysis proved to be one of the best methods for handling ground reaction data because of its ability to smooth the data and also perform other statistical analysis after converting it in the form of functional data.