Design and Implementation of Rehabilitation Training and Positioning System Based on Multi-Sensor Information Fusion

Towards human motion intention recognition in active rehabilitation, an algorithm of pattern recognition and localization is proposed. Large changes in the overall action signal were paid more attention rather than accuracy of local signal. The ARM of Cortex-M3 core was used in data acquisition and the Lab-VIEW to program human-computer interaction interface. Meanwhile the technology of intelligent action decision based on Neural Networks Artificial (ANN) was used and the wavelet packet was used to extract the signal feature of actions. Combined with the rehabilitation staff attributes information such as age, gender, stride length of walk, run, down stairs and up stairs to get the count of the staff's actions and location. Experimental results show that the performance of this method is 100% tested by genetic parameters optimization and the accuracy rate is 96.6667% by ROC. The target can be located according to the track of the target.