Research On Dyskinesia Assisted Diagnosis And Recovery System Based On Gait Analysis

Gait refers to the walking movement of human body under the control of the nervous system,which is coordinated by the motor system.Diseases that cause abnormal gait include Parkinson,stroke,hemiplegia,and cerebellar ataxia.With the increase of the elderly population in our country,the abnormal gait caused by neurological diseases and physical injuries is also increasing.Gait analysis can quantitatively assess the physical condition of patients,and plays an important role in disease diagnosis and recovery.Manual gait analysis requires professional medical knowledge,and the analysis results may be guided by subjective views.Existing intelligent gait analysis methods have high requirements for gait data acquisition equipment,and have many shortcomings.In order to solve the problems above,this thesis uses two-stream CNN to classify gait videos,and improves it by the difference between walking and other actions.For spatial stream,the key frame extraction algorithm is introduced to improve the single frame classification result;for temporal stream,inspired by the self-attention,this thesis used non-local block to extract the temporal long distance dependence to improve the classification effect.Considering that the pose estimation algorithm has achieved high accuracy,and equipment for acquiring depth images such as Kinect is widely used,skeleton points in 3D space is modeled for abnormal gait classification.on the basis of the related algorithms in 2D space,the motion data is modeled by spatial temporal graph and processed by graph convolution,and the adaptive graph convolution is introduced for optimization.It combines the body structure and the temporal motion feature to improve the classification effect of abnormal gait.Finally,based on algorithms above,this thesis designs and implements an abnormal gait disease diagnosis and recovery system.The system combines analysis algorithms and model modification to provide doctors and patients with gait analysis,diagnosis reference,gait parameters calculation,and result summary.The test shows that the system has reached the design goal in function,performance and algorithm.