EVALUATION OF GAIT BEHAVIOR WITH STATE SPACE VECTORS FOR THE DETERMINATION OF NEURODEGENERATIVE DISEASES

Abstract

Walking is a set of movements that take place with the control of the motor nervous system. Since neurodegenerative diseases caused by motor control disorders change the gait dynamics of individuals, analysis of gait signals is used in the detection of such diseases. In this study, it is considered that walking behavior can be evaluated from a systematic perspective to reveal the changing characteristics of gait. With this motivation, it has been proposed to evaluate gait signals with a state space approach that has not been used in previous studies. In accordance with this approach, walking behavior is evaluated with state vectors obtained in state spaces of different dimensions. The state spaces were formed for the left foot, right foot and the whole system, it was also studied with spaces that enable the left and right feet to be evaluated together. New series were obtained by calculating the size of the state vectors obtained from each state space and some statistical features were extracted from these series. In order to see the success of the proposed in differentiating neurodegenerative diseases from healthy ones, classification was done with the K Nearest Neighbor (KNN) algorithm by using ten fold cross correlation validation. The results reveal that state vectors are efective in distinguishing neurodegenerative diseases, with 90.5% accuracy. It is considered that the proposed approach may yield useful results in the analysis of other physiological signals.