Research On Abnormal Gait Detection Of Parkinson’s Patients Based On Vision

Parkinson’s disease Parkinson’s disease is a progressive nerodegenerative disease that often occurs in middle-aged and elderly people.Its main clinical manifestations are a variety of movement disorders,such as freezing gait.For Parkinson’s treatment,drug treatment is mainly used in the early stage,while for patients in the middle and late stages or severely affecting the quality of daily life,deep brain stimulation will be used for treatment.No matter what treatment plan is used,the follow-up patient’s movement disorder state will be tracked.Observation is very important.Nowadays,manual evaluation methods are commonly used.These methods require doctors and patients to invest a lot of time and effort,and the final results are also affected by artificial subjective factors.Therefore,automatic and objective motion evaluation through computer vision is for Parkinson’s patients.The rehabilitation evaluation is of great significance.This study started from the automatic assessment of the motion status of Parkinson’s patients by computer vision,and based on the timed up and go(TUG)video,explored two kinds of abnormalities in Parkinson’s patients through computer vision.The method of gait assessment.First,a lightweight computer vision method is proposed to realize the automatic recognition of the patient’s shuffling step.Shuffling step is a common form of gait disorder and an important factor used by doctors to judge the patient’s treatment status.This method uses a three-dimensional convolution module to extract spatiotemporal features from the preprocessed TUG video,and then extracts and fuses the obtained features on different spatiotemporal scales,and finally classifies the obtained features to obtain the corresponding recognition results.This method mainly It is hoped that it can be applied to telemedicine,so that patients can evaluate their abnormal gait only by recording TUG video with a video recording device in an environment with few restrictions.In addition,we also explored a method to estimate the key points of the three-dimensional human body through two-dimensional images from multiple perspectives.Using the key points of the three-dimensional human body can help doctors more accurately know the patient’s movement at any time during the exercise test.The data enables doctors to objectively evaluate the patient’s movement status.Moreover,the detection of three-dimensional key points realized by the computer vision solution is more economical and convenient than the sensor-based solution.At the same time,in order to reduce the work for doctors to use these computer vision algorithms,a Graphical User Interface(GUI)is also designed to integrate these algorithms.Doctors can obtain the analysis results of these algorithms with simple button operations.