基于分步机器学习的智能假肢步态识别  

A locomotion-mode recognition method for intelligent prosthesis based on step by step machine learning

作　　者：刘磊 杨鹏 刘作军 LIU Lei;YANG Peng;LIU Zuojun(School of Building Environment Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002;School of Artificial Intelligence and Data Science,Hebei University of Technology,Tianjin 300130)

机构地区：[1]郑州轻工业大学建筑环境工程学院,郑州450002 [2]河北工业大学人工智能与数据科学学院,天津300130

出　　处：《高技术通讯》2024年第11期1200-1210,共11页Chinese High Technology Letters

基　　金：国家自然科学基金(61803843);河南省教育厅重点项目(22B413012);河南省高等学校重点科研项目(24A480010)资助。

摘　　要：为了改善当前智能假肢步态识别在特征值选取和使用单一分类模型进行步态识别的不足,提高智能假肢穿戴者步态识别准确率,选择表面肌电信号(sEMG)作为步态识别信号源,提出基于灰度模型系数的特征提取方法,建立基于分步机器学习的步态识别模型。该模型选取灰度模型系数作为输入特征值,运用深度瘠波神经网络对步态做出初步识别,区分出易混淆步态,最后使用基于花授粉算法优化极限学习机参数的方法对易混淆步态作进一步的区分。实验结果表明,该方法对于平地行走、上楼、下楼、上坡、下坡、起立和坐下7种步态的识别准确率为98.25%,识别时间为70.48 ms,识别准确率高于单一机器学习模型。In order to remedy the shortcomings of feature value selection and single classification model in gait recognition of intelligent prosthesis,and improve the accuracy of gait recognition of intelligent prosthesis wearers,surface lectromyography(sEMG)signal is selected as the gait recognition signal source,a feature extraction method based on gray scale model is proposed,and a road condition recognition model based on step by step machine learning is established.The gray-scale model coefficient is selected as the input characteristic value.The deep fine wave neural network is used to identify the road condition and distinguish the easily confused gait.An extreme learning machine optimized based on pollination algorithm is used to further distinguish the confusing road conditions,and the final recognition accuracy is 98.25%and the recognition time is 70.48 ms.Compared with a single machine learning model,the accuracy of this method is higher than that of a single machine learning model for seven gaits:level ground walking,upstairs,downstairs,uphill,downhill,standing up and sitting down.

关 键 词：智能假肢 步态识别 表面肌电信号(sEMG) 灰度模型 极限学习机 

分 类 号：R318[医药卫生—生物医学工程] TP181[医药卫生—基础医学] TN911.7[自动化与计算机技术—控制理论与控制工程]