Research On Key Technologies Of Multimodal Fusion Rehabilitation Information System

As a high incidence disease of the elderly in today’s society,"stroke" has a very high incidence rate and mortality rate.Most of the elderly who survive have problems such as limb movement dysfunction,which requires long-term professional rehabilitation training to gradually restore normal limb movement ability.By collecting various signals generated by patients during exercise,the current movement status of the patient can be analyzed and judged,and combined with human motion ability assessment technology,it can help the patient achieve the best recovery state.Most stroke patients may experience sequelae of impaired ankle joint mobility after surgery.The structure of the ankle joint is complex,and its recovery process is long and difficult.The evaluation of ankle joint mobility can directly affect the patient’s recovery process.At present,the evaluation of ankle joint movement ability in clinical practice is mostly based on traditional methods such as evaluation scales or subjective evaluations by rehabilitation therapists.However,the number of rehabilitation therapists in China is very scarce,and relying on therapists to evaluate ankle joint movement ability will face enormous pressure.Therefore,this article explores new evaluation methods from the perspective of neural networks,which not only improves the accuracy of evaluation but also greatly liberates manpower,Ensure the efficiency and quality of ankle joint recovery.Firstly,this article discusses the current research status of multimodal information fusion and human motion evaluation methods both domestically and internationally,laying the foundation for the development of the paper.Secondly,regarding the research object of this paper,the ankle joint,the physiological structure and movement form of the ankle joint were first analyzed,and a motion model and evaluation model of the ankle joint were established through three muscle contraction modes.The evaluation features and types of collected signals were proposed.Subsequently,a multi-source sensor information collection system was constructed based on the features involved in the evaluation model.The selection and position distribution of force sensors and electromyography sensors,as well as the hardware design of the information collection system,were introduced.Based on this,corresponding information collection work was carried out,and the collected signals were preprocessed such as filtering and noise reduction.Finally,the reliability of the information collection system was verified through experiments.Then,feature extraction and fusion of multimodal information were carried out,and the features of sEMG signals were analyzed in both time and frequency domains.Complementary filtering fusion algorithms and Kalman filtering were used to fuse the joint force,joint motion angle,and surface electromyography signals generated during ankle joint movement.Multimodal information fusion can be used to evaluate the motion ability of the ankle joint,Avoid erroneous evaluations caused by external factors such as sensor measurement errors.Finally,experiments were conducted and patient motion data was collected,and it was determined to use joint degrees of freedom to evaluate ankle joint motion ability.The BP neural network evaluation model was trained and validated,and the accuracy of the evaluation model was verified through a dataset.Compared with traditional evaluation methods,the superiority of the BP neural network evaluation method was demonstrated.