Impact of training on long-range correlation of gait cycle

In order to study the difference of gait cycle characteristic between ordinary people and trained people, human gait signal was measured by an electrostatic sensor instead of accelerometer. Self correlation algorithm was applied to obtain the gait cycle series, further to calculate the increment series. The increment series was decomposed into magnitude series and sign series. A detrended fluctuation analysis algorithm was applied to analyze these series for 5 subjects without training and 5 others trained. It is found that the magnitude series of trained subjects show stronger long-range correlation than those without training, and the sign series of trained subjects show stronger long-range anti-correlation than the other.