Authors

Abstract

One of the essentials of intelligent prosthetics design is to recognize the wearer's movement intention, to provide the wearer with the corresponding control strategy and movement assistance. The 11 independent gait patterns and 5 transformed gait patterns are recognized by the self-designed human lower limb motion data measurement system. The human gait pattern is classified by the linear discriminant analysis (LDA) classifier, and the recognition accuracy is evaluated by K-fold Cross Validation(K-CV). The average recognition accuracy of independent gait patterns is 90.91%. In the independent gait pattern, the lowest recognition accuracy of DS1 gait phase is 90.53%, and the highest recognition accuracy of SS2 gait phase is 91.36%. The overall average recognition accuracy of the transformed gait pattern is 92.67%, the lowest recognition accuracy of DS1 gait phase is 91.93%, and the highest recognition accuracy of SS1 gait phase is 93.31%. The main reason affecting the recognition accuracy is that some gait patterns have similar motion characteristics. The method proposed in this study can accurately predict the wearer's locomotion mode and serves as a reference for gait pattern recognition, prediction, and control strategies in intelligent prosthetic devices.

Abstract in Chinese

智能假肢设计的关键之一是识别佩戴者的运动意图，为佩戴者提供相应的控制策略和运动辅助。通过自主设计的人体下肢运动数据测量系统对11种独立步态模式和5种变换步态模式进行识别。采用线性判别分析（LDA）分类器对人体步态模式进行分类，并采用K-fold交叉验证（K-CV）对识别精度进行评价。独立步态模式的平均识别准确率为90.91%。在独立步态模式下，DS1步态相位的识别准确率最低为90.53%，SS2步态相位的识别准确率最高为91.36%。变换步态模式的整体平均识别准确率为92.67%，DS1步态相位的最低识别准确率为91.93%，SS1步态相位的最高识别准确率为93.31%。影响识别精度的主要原因是某些步态模式具有相似的运动特征。本文提出的方法可以准确预测佩戴者的运动模式，为智能假肢的步态模式识别、预测和控制策略提供参考。