AUTHOR=Xu Dongfang , Wang Qining 

TITLE=On-board Training Strategy for IMU-Based Real-Time Locomotion Recognition of Transtibial Amputees With Robotic Prostheses

JOURNAL=Frontiers in Neurorobotics

VOLUME=Volume 14 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2020.00047

DOI=10.3389/fnbot.2020.00047

ISSN=1662-5218

ABSTRACT=This paper puts forward an on-board strategy for training model and develops the real-time human locomotion recognition study based on the trained model utilizing two inertial measurement units (IMUs) of robotic transtibial prosthesis. Three transtibial amputees were recruited as subjects for the study to finish five locomotion modes (level ground walking, stair ascending, stair descending, ramp ascending and ramp descending) with robotic prostheses.
An interactive interface was designed to collect sensors' data and instruct to train model and recognition. In the study, the analysis of variance ratio (with no more than 0.05) reflects the good repeatability of gait. Based on the on-board trained models, the subjects finished the five locomotion modes for real-time recognition. The on-board training time for SVM, QDA and LDA are 89 s, 25 s and 10 s based on a 10000$\times$80 training data set, respectively. It costs about 13.4 ms, 5.36 ms and 0.067 ms for SVM, QDA and LDA for each recognition. The recognition results indicate that SVM classifier can achieve the highest recognition accuracy (98.62%) while QDA and LDA classifiers achieve 96.78% and 93.75%, respectively. Taking the recognition accuracy and time consumption into consideration, we choose QDA for our next study. The mean recognition accuracy of the three subjects are 97.19% by QDA, and it can achieve more than 95% recognition accuracy for each locomotion mode. The study provides a preliminary interaction design for human-machine (prosthesis) in next clinical application. The study just adopts IMUs not multi-type sensors fusion to improve the integration and wearing convenience, and maintains comparable recognition accuracy with multi-type sensors fusion at the same time.