Find in Library

Problem arises when distinct morphologic changes are not seen in the electromyographic waveform of normal (control) and Parkinson's subjects during data interpretation. This study aimed to ascertain whether neuro-degenerative disease, e.g., Parkinson's disease (PD) affects gait and mobility with comparison to the healthy control. Fifteen subjects (left and right foot) from both the groups are selected where the signal is obtained using force-sensitive resistors (Gait Dynamics in Neuro-Degenerative Disease Data Base). The proposed methodology is divided into five parts: (i) 1 hr recording of gait dynamics data are segmented into three intervals (0-20 min, 20-40 min and 40-60 min); (ii) Normalization of each segmented data (20 min), i.e., preprocessing (noise and baseline drift removal); (iii) Then the frequency domain powers for each segments are calculated which further introduced features in the successive stages for classification and clustering; (iv) The classification of Parkinson's disease and healthy control group is accomplished using Linear Discriminant Analysis (LDA); (v) Clustering of these two classes is performed using K-means clustering algorithm taking same sets of features. Certainly the classification and clustering results signify the classification probability using frequency domain power of gait dynamics/electromyogram signal. The re-substitution error of LDA during classification is found to be 21.11%. Moreover, significant and precise classification and clustering results are achieved between PD and control taking left-right foot frequency domain power as classification features.