Adaptive detection of Ahead-sEMG based on short-time energy of local-detail difference and recognition in advance of upper-limb movements

Abstract

ObjectiveTo analyze the surface electromyography signal fragment ahead of the human movement (called Ahead-sEMG, the length of Ahead-sEMG is called the electromechanical delay (EMD)) and realize the advanced recognition of upper-limb movements. MethodsThe short-time energy method based on local-detail difference is proposed to adaptively detect Ahead-sEMG, and the AdaBoost classifier and VGG16 are employed to make the advanced recognition of upper-limb movements successful in this paper. ResultsFor each movement, there is at least one muscle channel whose sEMG is ahead of the movement. 7.01 %, 74.17 % and 18.82 % of EMD are greater than 0 ms and less than 30 ms, greater than 30 ms and less than 150 ms, and greater than 150 ms, respectively. The average length of EMDs is 101.52 ms. The EMDs of the brachioradialis, biceps and superior trapezius are 86.03 ms, 86.37 ms and 127.69 ms, accounting for 22.22 %, 16.67 % and 61.11 % of the longest EMD channels. The average EMD length of men is 45.32 ms, and that of women is 73.81 ms. AdaBoost and VGG16 are employed to classify, with an average recognition rate of 96.91 %, 70.73 % and 81.15 %, respectively. ConclusionThe experimental results show that the Ahead-sEMG is extracted precisely by the proposed method; the general time of the sEMG ahead of the action is 30–150 ms; the length of the Ahead-sEMG is related to muscle channels for the production of movement; the Ahead-sEMG can perform advanced movement recognition; and the study of the Ahead-sEMG is of great significance for movement advance recognition and further understanding of brain intention.