A method for automatic recognition and analysis of hemiplegic gait based on machine learning

Abstract

A method for automatic recognition and analysis of hemiplegic gait based on machine learning, including gait feature extraction, gait feature recognition, feature importance ranking, and Bayesian result classification. The specific steps are as follows: (1) Based on the Kinect sensor, capture the three-dimensional coordinate position of the subject's skeletal nodes; (2) Use the Euclidean distance algorithm and the segmental center of mass algorithm to calculate the human body's center of mass position movement range, stride, pace and other spatio-temporal characteristics ; (3) Establish the mapping relationship between the gait feature set in the input space and the corresponding markers in the output space; (4) Use the fuzzy binary comparison decision method based on information gain to rank the importance of feature combinations; (5) Model performance analysis. The method of the invention can obviously reduce the error rate of the doctor's subjective judgment on the patient's condition degree, and provide auxiliary data and evaluation results for clinical rehabilitation doctors.

Description

technical field [0001] The invention relates to a method for automatic identification and analysis of hemiplegic gait based on machine learning, and belongs to the technical field of auxiliary clinical diagnosis. Background technique [0002] Impairment of walking is one of the functional impairments that hemiplegic patients urgently need to recover. In clinical rehabilitation, quantitative gait information obtained by analyzing hemiplegic gait can provide a basis for revealing the cause of abnormal gait, correcting abnormal gait, formulating rehabilitation treatment plan, and evaluating the effect of rehabilitation intervention. [0003] Most clinicians use subjective observation and scale scoring in the gait assessment of hemiplegic patients, but observation and assessment with individual subjective differences are not reliable enough in clinical treatment, and scale scoring provides Information about the patient's daily exercise capacity is also often considered imperson...

AI Technical Summary

Problems solved by technology

[0004] In 2011, Tsinghua University developed a high-speed camera. Its core technology is to paste reflective markers on the bone nodes of the subject, collect the position of the reflective markers through an infrared camera, and extract dynamic gait information based on later image processing. This method requires complex image processing work, which is not conducive to popularization

However, this method lacks the result processing of the obtained dynamic characteristics and spatiotemporal characteristics, and does not rank the importance of variables in the characteristics, which is not convenient for disease diagnosis and optimization of rehabilitation programs.

Images