Research on Surface EMG Based Accurate Perception Method for Exoskeleton Robot Control

For coordinating and high-precision control of the lower limb wearable exoskeleton, surface electromyography (sEMG) which reflected the neuromuscular activity is chosen as the main signal source to obtain more accurate motion pattern in this paper. At first, 4-channel sEMG signals which can be described separately as biceps femoris, vastus medialis, rectus femoris, and gastrocnemius are collected and de-noised using wavelet transform (WT) algorithm. And then following the multi-scale decomposition, the singular value of wavelet coefficient can be extracted to construct the feature vector which will be the input of pattern recognition. In the mean time, a least squares support vector machine (LS-SVM) classifier is investigated to classify different movement patterns. Finally, six movement patterns (downhill, running, squatting, standing, upslope, walking) are successfully identified. Experiments show that the proposed method performs a high accuracy with fewer data samples and provides a great potential in the practical application of wearable exoskeleton control with sEMG.