CAJ | 학술논문

表面肌电信号（sEMG）被广泛应用于假肢控制，但是由于解剖组织、生理状态等因素使肌电信号表现出较大的个体差异，传统的参数归一化方法需要大样本的训练构建适合用户的肌电控制模型，新近发展起来的一种基于动作因素和个体因素的双线性模型为肌电假肢控制提供了新的思路。利用指总伸肌的综合强度和肌电活动的空间分布特征构建双线性模型，将指总伸肌肌电活动从神经-肌肉生理机制的角度分解为用户相关的个人因素矩阵和与动作相关的动作模式矩阵。为了验证该模型，设计了食指、中指、无名指在20％ MVC 、40％ MVC、60％ MVC 3个力量水平的单指按压实验，然后利用32通道柔性阵列电极所采集的6名志愿者的指总伸肌sEMG完成双线性模型的构建，在较少的样本训练下实现了对力量水平和手指模式的较好识别。结果表明双线性模型可用于简化肌电假肢接口的训练过程，对肌电假肢手力量和手指动作控制有较大的应用前景。
표면기전신호（sEMG）피엄범응용우가지공제，단시유우해부조직、생리상태등인소사기전신호표현출교대적개체차이，전통적삼수귀일화방법수요대양본적훈련구건괄합용호적기전공제모형，신근발전기래적일충기우동작인소화개체인소적쌍선성모형위기전가지공제제공료신적사로。이용지총신기적종합강도화기전활동적공간분포특정구건쌍선성모형，장지총신기기전활동종신경-기육생리궤제적각도분해위용호상관적개인인소구진화여동작상관적동작모식구진。위료험증해모형，설계료식지、중지、무명지재20％ MVC 、40％ MVC、60％ MVC 3개역량수평적단지안압실험，연후이용32통도유성진렬전겁소채집적6명지원자적지총신기sEMG완성쌍선성모형적구건，재교소적양본훈련하실현료대역량수평화수지모식적교호식별。결과표명쌍선성모형가용우간화기전가지접구적훈련과정，대기전가지수역량화수지동작공제유교대적응용전경。
Surface electromyography (sEMG)has been widely used as the control source in hand prosthesis;however,some factors, such as the individual anatomy property and physiological state can cause different sEMGs even in the same motor task.Traditional pa-rameter normalization is usually employed to construct a general sEMG model among individuals,which requires extensive sample train-ing.To cope with this problem,the bilinear model representing the activities with motion factors and individual factors provides a novel approach for myoelectric prosthetic control.The bilinear model is introduced to extract the neuromuscular characters of the extensor digi-torum’s activities.Firstly,the activities of the muscle are described with the multi-dimensional feature vectors including integrated sMEG amplitude and spatial distribution.Secondly,the vectors are decomposed into user-dependent matrices and motion-dependent ma-trices.To test the model,6 subjects were recruited for finger force-tracking tasks.The finger experiments on the index,middle and ring fingers with the force levels of 20%MVC,40%MVC and 60%MVC were designed.At the same time,the sEMGs were collected from the extensor digitorum with a 32-channel flexible electrode array.Then,the bilinear model was built to recognize the finger pattern and force levels under less sample training condition.The experiment results show that the proposed bilinear model can classify the finger pattern and force levels through only a few interactions,which suggests that the bilinear model can simplify the training procedures of my-oelectric prosthetic interface,and has good application prospect in the control of the finger motions and force levels of myoelectric pros-thetic hand.