Data smoothing, filtering and processing

Despite increasing improvements in the precision of modern-day experimental equipment, biomechanical assessments of human movement are almost always influenced by one or more sources of error (otherwise known as noise). Of the many error sources, the ones introduced during the digitisation of motion capture data and those that are caused by environmental factors, skin movement artefacts and other electrophysiological signals (e.g. the electrocardiogram) are amongst the most common. Although these error sources are widely recognised, they can significantly affect the quality of the data if their influence is not appropriately negated by applying an algorithm that smooths (or filters) the data. This section provides an introduction to a number of different methods that can be used to smooth noisy biomechanical data. These methods include the relatively simple but inflexible Hanning algorithm, as well as more complex and flexible methods, such as the Butterworth technique and the cubic spline and quintic spline algorithms. Examples are also provided to highlight the effect of applying these different techniques to a time series of noisy kinematic data collected during walking.