INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 02, FEBRUARY 2020 ISSN 2277-8616 6250 IJSTR©2020 www.ijstr.org Classification Of Gait Dynamics In Neurodegenerative Disease Patients Using Machine Learning Techniques S.A.Vajiha Begum, Dr.M.Pushpa Rani Abstract: The key objective of this paper is to classify the gait dynamics of Neurodegenerative patients using Machine Learning (ML) techniques. The studies state that every individual has a unique walking style and these patterns are termed as gait. Human gait analysis recognizes people from the way they walk, which replicate the individual’s unique movement pattern. The severe walking abnormalities in human are caused by progressive brain dysfunction. Accurate diagnosis of the particular brain disorder helps to start early treatment procedures. Lack of personal analysis of the patient to identify the brain disorder needs machine learning based techniques to accurately diagnose and classify the Neurodegenerative Diseases (NDD) such as Amyotrophic Lateral Sclerosis, Parkinson’s, Huntington’s Diseases and the healthy control subjects. The Recurrence Quantification Analysis (RQA) and Fast Walsh-Hadamard Transform is implemented to quantify the input gait signals to extract the gait parameters. The extracted parameters are further used for classification of NDD diseases with the multi SVM and Random forest algorithm gives 89% and 91% accuracy. Hence, it is possible to classify the normal persons and persons with brain disorders by the Machine Learning techniques with the gait dynamics. Index Terms: Gait Dynamics, Neurodegenerative Diseases (NDD), Recurrence Quantification Analysis (RQA), Fast Walsh-Hadamard Transform, Multi SVM, Random Forest, Machine Learning. ——————————  —————————— 1. INTRODUCTION Millions of People are affected by Neurodegenerative diseases worldwide. The most common Neurodegenerative diseases such as Amyotrophic Lateral Sclerosis (ALS), Huntington’s and Parkinson disease will reflect in abnormal and slow walking movement in humans. The Neurodegenerative disease is caused due to the loss of neurons and axons in the central nervous system which leads to serious brain disorder. The persons with brain disorders will show symptoms like tremor, slowness of walking or loss of walking, shaking in arms, hand, legs, face and jaw and muscle inflexibility [1]. The study shows that each and every individual has a different walking patterns and these patterns are termed as gait [2]. There is difference in the gait dynamics of healthy individual and the persons with neurodegenerative diseases. Mostly aged peoples are unaware and affected with Neurodegenerative diseases. The exact brain disorder is unable to diagnosis by the physicians and hence require accurate diagnosis method to start the treatment process earlier.The automatic diagnosis of particular neurodegenerative disorder with gait pattern using the machine learning technique is proposed to give efficient and accurate classification of Amyotrophic Lateral Sclerosis (ALS), Huntington’s, Parkinson disease and healthy controls without human intervention [10]. The input gait signals used in this work is taken from the physionet public database. The gait signals are non-linear, non-stationary and recurrence in nature. To deal with the non-linear data, Recurrence Quantification Analysis (RQA) and Walsh-Hadamard Transform is applied to measure the input gait signals for the extraction of gait parameters. The RQA quantifies the recurrence of gait time series. The recurrence nature of gait is constructed using the Recurrence plot and features such as Recurrence Rate (RR), Determinism, Entropy and Average Diagonal Length are extracted for further classification. The Walsh-Hadamard Transform is also applied with the gait signals to gather the statistical parameters. These extracted parameters are classified with Multi SVM and Random forest classifiers. 2. LITERATURE REVIEW Kartikay Gupta et al., [4] presented an efficient classification method and generated a new set of features with the autocorrelation and cross correlation between the gait time series. Mutual Information (MI) analysis is applied and generated 500 features for classification. A rule-based classifier technique using single decision tree classifier was implemented for classification of Huntington’s Disease (HD), Parkinson’s Disease (PD), Amyotrophic Lateral Sclerosis (ALS), and Neurodegenerative Diseases (NDD) with 500 features achieved classification accuracy of 88.5%, 92.3%, 96.2%, and 87.5%. The validation for HD vs control, PD vs Control, ALS vs control and NDD vs. control got accuracy of 80%, 80%, 90% and 73.33%. P.Prabhu et al., [6] proposed a efficient method to quantify the non linear gait data using Recurrence Quantification Analysis (RQA). The parameters gained from RQA is used to determine the periodicity, complex behaviour and deterministic nature to describe the individual gait and improved the binary classification accuracy using SVM and PNN gives 96% and 100 % result. Pushpa Rani et al., [9] presented a survey result on gait pattern classification and recognition by comparing Wavelet Descriptor with ICA and Hough transform with PCA. For the classification purpose SVM and nearest neighbour classifiers are used. The gait recognition is done by cumulative match scores and results were finally shown. Pushpa Rani[5] proposed a modified version of Extreme Learning Machine called Hybrid Extreme learning Machine (HELM). The Analytical Network Process ————————————————  S.A.Vajiha Begum, Research Scholar, Department of Computer Science, Mother Teresa Women’s University, Kodaikanal. TamilNadu, India, E-mail: vajihabegum7391@gmail.com  Dr M.Pushpa Rani, Professor, Department of Computer Science, Mother Teresa Women’s University, Kodaikanal. TamilNadu, India, E-mail: drpushpa.mtwu@gmail.com