Research On Sequential Upper Limb Movements And Robot Control

Taking human motion state as input signals to control peripheral equipment,such as a robot and a mechanical arm,is one of the major subjects in the field of Human Computer Interaction(HCI).Online recognition of different motion patterns can help users to interact,in a more natural way,with machines by doing simple and flexible limb movements.In this paper,the robot control technology based on the detection of human motion states was taken as the study object,and detection and recognition of surface electromyography(sEMG)signals were taken as the technological means.The study is to investigate the sequential relationship and recognition algorithm of upper limb movements,with taking into two aspects: offline analysis and online control,which realized the robot control technology based on detecting the sEMG signals of motion state.Contributions of the current study include:Offline analysis of sequential upper limb movements based on sEMG signals.For single-feature recognition,two features of the sEMG signals,named mean absolute value(MAV)and waveform length(WL),were extracted.The 6-class sequential upper limb movements were successfully classified by using interval segmentation method,decision-tree and pruning algorithm,and the classification accuracy of MAV and WL reaches 95.24% and 94.05%,respectively.After that,multi-feature fusion recognition was carried out by using the techniques of feature-level fusion and decision-level fusion,with the accuracy reaching 96.43% and 98.81%,respectively.Experimental results show that the classifier developed in this study can do well in offline recognition of movement as required.HCI of sequential upper limb movements based on online sEMG signals.Online control on NAO(a humanoid articulated robot)was conducted by using synchronous paradigm and asynchronous paradigm,respectively.In synchronous paradigm,the follow-up control strategy was used,which allowed the NAO robot to follow up the motions of the subject.The synchronous paradigm provides an idea for rehabilitation therapy.In asynchronous paradigm,the mapping control strategy was used,and the above sequential movements were taken as the state switching condition to switch NAO's motion state.The above experimental results show that the detection of sequential movements based on sEMG signals is effective to realize the function of HCI.Primary research on natural motions of upper limbs based on sEMG signals.To further make clear of the natural motions of human limbs and recognition algorithms,we have generalized the 6-class required motions,as mentioned above,to 3-class human natural motions,and made classification and recognition by using the methods for analyzing the required motions.Experimental results suggest that there is a sequential relationship behaved by upper limb's joints when capturing a target,but such sequential relationship is not as significant as that in the task of the fixed sequential movements.Decoupling the coupled sequential relationship will be the key to recognize human natural motions.In conclusion,this study has achieved the offline analysis and online control on sequential upper limb movements based on sEMG signals,which has a prospect of application in fields of medical rehabilitation and weapons,etc.