基于肌音信号的步态动作模式识别研究

通过采集腿部肌肉5个通道的肌音信号,利用3层决策树对跑步、上楼、下楼、走路、静止5种步态动作进行模式识别研究。在决策树的第1层和第2层,应用双阈值门限法识别静止和跑步两种步态模式,在第3层,提出基于步态信号的自适应不等长分割算法以及改进的模糊熵算法,利用线性分类器对走路、上楼、下楼进行分类识别。结果表明:双门限阈值法可有效地对静止和跑步进行识别,当采用改进的模糊熵特征时,对走路、上楼、下楼3种步态模式的分类准确率达到了94.87%;而当综合利用近似熵、样本熵和改进的模糊熵3种特征时,其分类准确率达到了98.76%。

Discrimination of Gait Patterns Based on Mechanomyographic Signal

Five gait patterns, walking, going upstairs, going downstairs, standing and running, were distinguished by five-channel mechanomyographic(MMG) signals of leg muscles with hierarchical decision tree. Running and standing were recognized with two thresholds method in the first and second layer of the decision tree. In the last layer of decision tree, an unequal division algorithm and a modified fuzzy entropy algorithm were proposed to extract the features of walking, going upstairs and going downstairs. Then the three gait patterns were distinguished by a linear classifier. Experimental results showed that two thresholds method could recognize standing and running entirely. The other three patterns could be classified with overall accuracy rate of 94.87% when only modified fuzzy entropy was used as the feature, and 98.76% when approximate entropy, sample entropy and modified fuzzy entropy were used simultaneously.