Design And Implementation Of Gait Analysis System Based On RGB-D Information

Gait analysis system,as a diagnostic and monitoring instrument,is often used in the fields of sports and rehabilitation in clinical medical treatment.Due to expensiveness and inconvenience of the large gait analysis system in hospitals and other public infrastructure,patients have great demand for personal solutions in low price.As a result of high requirement of space and medical equipment,there are few products that meet patients' needs.In consideration of the requirements of gait analysis with the progress of attitude estimation research,a solution is proposed,which integrates RGB information and depth information,and extracts 3d keypoint sequences through pose estimation network for gait analysis.Trading off between requirements of high-precision data and low-cost systems,functional and non-functional requirements of the system were summarized based on application scenarios and patient requirements,and basic system framework was designed.Firstly,under the condition of low cost and limited space,for making full use of information of multi-source data and modeling advantages of deep learning network,an additional 3d full-convolutional structure is applied based on the improved OpenPose,to generate 3d keypoints of recorded data in high accuracy.Secondly,ground fitting is used instead of camera calibration,to simplify the acquisition of mapping relationship between camera coordinate system and real space,for restoring keypoints into real space.Through division and normalization of walking direction,several straight walks were obtained,which improves the accuracy and robustness of gait parameter calculation.Thirdly,based on the needs of 3d keypoints extraction and 3d gait analysis,the existing data acquisition scheme is improved through framerate controlling,time alignment,down-sampling and parallelization.It can output steadier RGB-D image sequences,and acquire necessary information such as timestamp and camera parameters synchronously.Finally,visualization of global,sequential and dynamic gait parameters is realized in several different formats for data display.The RGB-D gait analysis system based on PC platform is integrated to realize highaccuracy gait analysis.Through testing,the gait analysis system basically achieves the expected functional and performance goals.