Research On Evaluation Method And Evaluation System Of Lower Limb Motor Ability Based On Multimodal Data

With the development of the economic level and the improvement of medical level,the number of elderly people in China has increased year by year,and the situation of aging has become increasingly severe.Most elderly people’s athletic ability declines with age,which undoubtedly brings various inconveniences to their daily lives and has a negative impact on the spiritual level.Exoskeleton equipment is an electromechanical device used for restoring and enhancing the athletic ability.Lower extremity exoskeleton walking equipment can help the elderly with reduced mobility during walking.Therefore,equipping the corresponding lower extremity exoskeleton walking device according to the evaluation results of the lower extremity athletic ability is the best method to solve the above problems.This thesis conducts research from the three aspects of multi-modal data collection and processing,the evaluation method of lower limb mobility and the evaluation system.The main contents of the thesis are as follows:Firstly,the definitions of the basic axial and joint motions of the human body are introduced,and the characteristics of lower limb movements are analyzed to pave the way for studying the gait characteristics of the walking process.VICON MX three dimensional gait system is used to collect multimodal data of the human body,including gait video,lower limb dynamics data and lower limb kinematics data.Gait contour images are extracted from the gait video by the improved Vi Be algorithm to remove the effects of environment and clothing,and statistical analysis of the changes of the lower limb joint angle and the ground reaction force is provided to provide theoretical support and data support for the establishment of lower limb motor function evaluation models.Secondly,this thesis proposes a lower limb motor function evaluation method based on the combination of improved convolutional neural network and decision fusion algorithm.The convolutional neural network is improved by introducing a spatial pyramid pooling layer and a COCOB optimization algorithm to further improve the recognition accuracy of the gait image.At the same time,the optimal classification is selected after extracting the knee joint angle and ground reaction force features by principal component analysis.algorithm.A decision fusion algorithm is used to establish an evaluation index of lower limb mobility based on the recognition probability of three modal data.Comparing the recognition performance and evaluation performance of different modal data,the results show that the best three-modal fusion result has reached a recognition accuracy of 96.15%,and the lower limb motor function evaluation indexes of the best single-modal,dual-modal and three-modal are significantly negatively correlated with GRAS-M,with correlation coefficients of 0.63,0.72,and 0.82,respectively.Furthermore,this thesis proposes a lower limb motor function evaluation method based on the combination of Xception-LSTM and kernel principal component analysis.The Xception-LSTM network model is used to extract the fully connected layer features of the gait video image and fuse them with the dynamic data and kinematics data of lower limbs in the feature layer.Then the fused features are subjected to dimensionality reduction by kernel principal component analysis,and an evaluation index of lower limb mobility is established from this.The experimental results show that this index can quantify the lower limb motor function,there is a significant statistical difference between people with different fitness levels,and it has a significant linear correlation with GRAS-M(r(28)0.90).Finally,based on the evaluation results of the Xception-LSTM model,a lower limb motor function evaluation system based on the Django framework is built.The evaluation system uses the Django framework and My SQL database to establish evaluation information management,information display,and role management modules,and realize the query,display,statistics and management of evaluation information.The lower limb motor function evaluation system is convenient for the subjects,assessors and system administrators,and it also provides data support for lower extremity exoskeleton walking equipment and elderly care.