Figure 3

Fourier and wavelet illustrations. The Fourier and wavelet power spectra for M5, the basic features of which are typical of all the individuals studied here, illustrate how movement predictability (MP) is revealed through intermittent movement autocorrelation. Panel (a) shows the Fourier spectrum computed using a Tukey (bell cosine) taper and smoothed with a modified Danielle smoother over the immediately adjacent frequency locations. For easy comparison against a random walk process that corresponds to movement with no MP, the power is normalized so that a white noise signal with variance equal to the variance in the observed net displacement time series has the constant power spectra of one across all frequencies. Spikes in power indicate significant periodic activity at 1, 2 and 3 cycles per day. Panel (b) shows contour plots of significant regions in the wavelet power spectrum, which allows time-localized determination of when MP is present, and shows that MP is not constant through time. Areas within closed curves (thick black lines) indicate movement that is significantly periodic.