基于多模态数据融合的脑性瘫痪儿童连续运动预测算法研究OA

Objective To accurately predict the continuous motion trajectories of children with cerebral palsy using multimodal data fusion and deep learning methods,and provide support for personalized rehabilitation training and the design of assistive devices.Methods By integrating surface electromyography(sEMG)and inertial measurement unit(IMU)data,a continuous motion prediction model based on a convolutional neural network(CNN)and long short-term memory(LSTM)architecture was constructed.The lower-limb sEMG and IMU data of 5 children with cerebral palsy during walking were collected and preprocessed by denoising and normalization.Spatial features were extracted via CNN,and temporal dependencies were captured by LSTM to construct a prediction model of knee joint angles.The model performance was evaluated using the mean absolute error(MAE)and coefficient of determination(R2).Results The evaluation results demonstrated that the model achieved high accuracy in predicting the gait trajectories of children with cerebral palsy,with MAE=0.083 7±0.012 8 and R2=0.986 1±0.007 1.The predictive performance of the model was effectively enhanced by the multimodal data fusion combined with CNN-LSTM architecture which accurately captured the complex motion patterns of children with cerebral palsy.Conclusion The motion prediction model based on multimodal data fusion and the CNN-LSTM architecture constructed in this study provides a new technical method for the quantitative assessment of motor dysfunction in children with cerebral palsy,and has potential application value,especially in the design of personalized rehabilitation programs and assistive devices.