Using feed-forward neural network for complex static balance signal characterization with chaotic features

In this paper largest Lyapunov exponent and low frequency power ratio of the Center-of-Pressure (COP) signal are calculated and considered as two discriminative features to characterize prolonged standing and its effect on postural control in elderly individuals in comparison to adults. It is unknown how elderly individuals behave during standing and how demanding such a task is for them. We recorded the center of pressure position of 16 elder subjects and 24 young subjects while they performed standing (30 second). Statistical and frequency-based features are extracted and an analysis is performed to find the most appropriate and discriminative features for elder and young subjects posture signals discrimination. Standard deviation and mean value as statistical parameters, low frequency power ratio as frequency based feature and largest Lyapunov exponent (LLE) of COP signal as the chaotic feature are computed to discriminate the COP signals of young and old subjects. LLE is estimated to show the impact of chaotic behavior in static balance relative to the age. Implementations on the normal subjects demonstrate that working in frequency and chaotic domains is preferred to time domain and statistical analysis and largest Lyapunov exponent of the posture signal can be representatively used for COP signal characterization.