The measurement of CoP trajectory has become the de-facto standard for posturography. However, in recent years, inertial sensors have been proposed as a portable alternative to force plates. Although their demonstrated potential, the functional interpretation of these methods remains limited, and no standard approach exists for inertial signal processing.
This work aims to analyse and compare GRF- and IMU-based metrics (obtained from a single IMU positioned on the trunk at L5 level) in characterising postural control performance in 21 healthy participants on varying surface (i.e. solid ground and three foams of different stiffness) and visual (eyes open/closed) conditions, concurrently analysing how results are affected by different filtering cut-off frequencies.
In line with existing literature, GRF signals were lowpass-filtered at 10 Hz, while IMU signals at 0.5 Hz, 3.5 Hz, 5 Hz (i.e. band-width containing 95 % of the signal power), and 10 Hz. Time- and frequency-domain postural parameters were extracted from GRF and IMU signals.
Correlations between GRF- and IMU-based metrics resulted weak to moderate (0<|ρ|<0.7). Both GRF- and IMU-based metrics showed increased postural oscillations on foam surfaces, but opposite behaviours in frequency, with no significant difference among different foam types. GRF-based metrics highlighted higher postural oscillations under eyes-closed conditions, especially on foam, whereas IMU-based metrics showed no significant change except for range and root mean square displacement in the medio-lateral direction that decreased with eyes closed (e.g., 5 Hz low pass filtered IMU signal: on foam, median root mean square, with eyes open [25th-75th], 0.08 [0.05–0.12]; with eyes closed, 0.04 [0.03–0.06]).
Although describing the same behaviour, GRF- and IMU-based metrics capture different aspects of postural control: based on the inverted pendulum model, GRF-based metrics describe the postural adjustments, while IMU-based ones the postural performance.