Characterization of gait dynamics using fractal analysis for normal and Parkinson disease patients

Gait dynamics of human appears to have a self-similar pattern and this self-similarity is studied using various non-linear methods. In this paper we employ three non-linear methods namely Hurst´s rescaled range analysis method, Higuchi´s fractal dimension method, Approximate entropy method to quantize the gait dynamics of people suffering from Parkinson diseases (PD). Also they are asked to perform `Dual task´ while walking, to have a better understanding of the connection between central nervous system and the limb movement. Gait dynamics of normal healthy people are also included in this study as control group. They are also put under study to perform dual task. This data set has been acquired from MIT-BIH data base. All these methods are capable to detect the difference between `Normal walking´ and `Dual tasking´ for both the groups. The fractal exponents for `Dual task´ however produce interesting findings for PD subjects, suggesting that this type of walk is more chaotic and is more disordered than control group with the same task. Amongst the three methods utilized in this work approximate entropy method appears to be more sensitive, for both the groups.