طبقه بندي الگوهاي كينتيكي راه رفتن انسان به دو گروه سالم و بيمار با به كارگيري شبكه عصبي و كاهش فركانس نمونه برداري

چكيده انگليسي :

Human Walking Classification In Two Healthy And Non healthy Groups Based On Kinetic Pattern Using Neural Network And Reduced Frequency Marzieh Mostafavizadeh m mostafavizadeh@ec iut ac ir 2011 10 23 Department of Electric and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranDegree M Sc Language FarsiSupervisor Dr Farid Sheikholeslam sheikh@cc iut ac irAbstractHuman walking is a completely complicated activity in which both musculoskeletal and neural systems of the bodyare involved Human aging strongly affect walking pattern so the risk of balance impairment is increased Inaddition most of elderly have the problem of osteoporosis so falling can cause serious fractures especially in lowerlimb joints such as ankle knee and hip As in most cases the general circumstance of elderly such as blood pressure heart rate oxygen saturation etc is not proper so restoration operation will be challenging and burden extra cost As the result of these matters it will be very important to identify balance impairment in gait patterns especially inelderly people In fact elderly classification in to two groups and recognizing the elderly who are disposed high riskof falling can prevent falling by applying walking aids or protective supports Gait pattern can be explored by using3 special kinds of parameters EMG signals kinematic and kinetic parameters EMG signals are most recorded bysurface electrodes and contain useful information about muscle activities Kinematic parameters includedisplacement velocity and acceleration of joints involved in walking such as angular linear velocity orangular linear displacement of the joints Kinetic parameters refers to forces and moments that are responsible forchanging body state during motion include moments and ground reaction forces GRF exerted to the feet As theonly contact region of the body during walking is the feet the distribution of these forces are very effective inhuman balance EMG kinematic or kinetic signals contain important information about gait pattern and humanbalance but these information cannot be directly explored by the physician However in the past decade mostresearchers are focused on kinematic parameters or combination of EMG and kinematic parameters recording ofthese parameters need costly equipments such as Gait Analyzer or Motion Capture with at least 6 infrared camerasof high sensitivity arranged around the path way of walking these equipments are rarely available in Iran labs andusually very expensive and costly Another common equipment is force plate or platform which include 4 or 6 straingauge or piezoelectric pressure sensors and used to obtain ground reaction forces and moments exerted to bodyduring motion In this paper we use a common six channel strain gauge force plate with a pre amplifier to obtainthree force components and three moment components in reference planes media lateral frontal and transverseplanes Kinetic data have a very irregular nature and simply noise polluted In this paper we normalize the obtaineddata and rearrange them in to special matrixes as the inputs of a 3 layer feed forward neural network with a hiddenlayer Actually we explore two major problem with kinetic data The first one is to reduce the practical samplingfrequency and the other is to classify kinetic patterns in two group of healthy and non healthy people We use bothneural network and Anfis model to solve these problems and then compare the results Keyword Human Walking Balance Impairment Classification Neural Network Anfis Network

مصطفوي زاده، مرضيه

طبقه بندي الگوهاي كينتيكي راه رفتن انسان به دو گروه سالم و بيمار با به كارگيري شبكه عصبي و كاهش فركانس نمونه برداري

تعادل , پارامترهاي كينتيكي , مدل آنفيس