Characterizing Temporal Behavior for Human Unipedal Balance

We investigate the data obtained with time as a person stands one-legged on a force plate to identify underlying characteristics of a healthy participant’s balance process. We use signal processing tools to quantify any temporal patterns in the data, and subjects can be classified based on such a profile. To begin, we filtered the medial-lateral and anterior-posterior force and jerk signals to reduce the influence of machine noise and focus the study on a fluctuation frequency band of 0 – 15 Hz. Moreover, our preliminary analysis suggests that such signals may have temporally evolving statistics which we track. With its ability to analyze signals over multiple time scales, a wavelet transform is appropriate for examining such data and is used to explore joint time-frequency behavior. Its use also allows us to decompose the data into several time series components, or modes, corresponding to different frequency bands, which sum to give the original signal. We attempt to classify temporal patterns found through correlations measured for the sequence of wavelet coefficients corresponding to each mode.