Study On Safety Monitoring Network Techniques During Walking Assisted By Dynamic Aids

The number of the people with dyskinesia of lower limbs is on the increase because of an aging social population and high occurrence of traffic accidents. With the development of medical rehabilitation and rehabilitation engineering technology, dynamic aids have played a more important role in rehabilitation training and functional restoration of kinds of limbs dyskinesia patients. However, lack of a reliable way to gain the determinable features and monitor moment security may cause quadratic injury when patients are conducting self-training, which may result in serious physical and mental injury.In order to solve the problem, in this thesis, a set of monitoring system with dynamic aids was designed to process safety analysis and monition in patients' rehabilitation training using neural network. Moreover, the most parameters which could represent stability in this dynamic aids monitoring system was filtered, and optimized parameter was chosen as safety features. What is more, the performance of optimized parameter in different models of the system was discussed. Firstly, software and hardware facilities of monitoring were constructed, in which multi-sensor information of system was calibrated with traditional linear method and ant colony algorithm. The result of error check displayed that calibration using ant colony algorithm could effectively decrease crosstalk interference and nonlinearity error in systematic survey. Considering safety, mechanics tipping testing was also designed and processed, and relevant experimental data was collected. We did discriminatory analysis of tipping threshold and models utilizing BP neural network.Original signals of multi-sensor, handle reaction vector (HRV), Walking Tipping Index (WTI), Virtual Center of Gravity (VCG) and etc., were used as input features of network for recognition. Optimal VCG fused with original signals was utilized as safety feature of system. Finally, we did dynamic aids combining braces test experiments of five males, collected experiment data of four models separately, analyzed safety features by trained network, and investigated the effect of functional electrical stimulation brace on walking aids.According to the results in this thesis, monitoring by safety neural network in dynamic aids monitoring system can discriminate aids tipping models, and decrease collapse crisis. In future, the monitoring and evaluation of dyskinesia patients' self- rehabilitation could be popularized and applied, which will assist the design of more advanced dynamic aids and artificial motor neural prosthesis system.