Human Abnormal Gait Recognition Based On Smart-phone

Gait is a posture when walking,and normal gait is the result of coordinated movement of the human body.Diseases,heredity,accidental injuries and other factors can cause gait changes,resulting in abnormal gait.Analysis,identification and evaluation of abnormal gait have important value in clinical diagnosis and rehabilitation.The analysis and evaluation of abnormal gait in traditional clinical medicine is often difficult to use in daily life by using expensive professional equipment.With the development of smart phone-based gait recognition technology,the identification and analysis of abnormal gait of human body has a more economical and convenient measurement method,but the problems that arise with it need to be studied.Through experimental design,this thesis uses the smart phone to collect normal and abnormal gait data of the human body.After pre-processing,feature extraction,classification and parameter estimation,quantitative analysis,and the theoretical model of abnormal gait are established to verify each other,so as to realize the qualitative identification and quantitative analysis of abnormal gait based on smartphone.The main contents of the full text are summarized as follows The main work of this thesis is as follows:1.Develop and design a gait data acquisition APP for Android based on the acceleration and rotation vector sensors built into the smartphone.For the problems of jitter and mobile phone orientation during data acquisition,the quaternion method is used to convert the relative coordinate system of the mobile phone into the absolute coordinate system of the earth,and the data preprocessing is performed by using the dual-tree complex wavelet algorithm.2.In the absence of quantitative analysis such as parameter estimation and quantitative analysis in the gait field,biomechanical parameters: time and space timespace parameters are used to study the step size,stride frequency,pace and angle;taking into account human gait Periodically,a gait template is constructed to extract the gait cycle sequence.An abnormal gait model based on the lower extremity of the human body,the pendulum model,was constructed and verified experimentally.3.In the qualitative analysis of abnormal gait,the extracted time-frequency features and partial time-domain features are taken as the original 36-dimensional features.Aiming at the problem that the feature dimension of the abnormal gait movement is complex,redundant and the correlation between the features is high,the effective features are selected,and the feature selection algorithm CFS is used for feature selection and data dimension reduction.In the design of the classifier model,different machine learning algorithms are selected.According to the characteristics of different classifiers,the optimal parameters are selected,the parameter optimization scheme is designed,the classifier indicators are evaluated,and the optimal classifier is selected.The classification recognition accuracy can reach 96%,and the classifier model can reach 88.5% after generalization.