Improving robustness against electrode shift of high density EMG for myoelectric control through common spatial patterns

• DOI: 10.1186/s12984-015-0102-9
• Publication Date: 2015-12-02T08:19:08Z
• Authors: Lizhi Pan; Dingguo Zhang; Ning Jiang; Xinjun Sheng; Xiangyang Zhu
• Keywords: Robustness

Most prosthetic myoelectric control studies have concentrated on low density (less than 16 electrodes, LD) electromyography (EMG) signals, due to its better clinical applicability and computation complexity compared with high (more HD) EMG signals. Since HD electrodes been developed more conveniently wear respect the previous versions recently, signals become an alternative for prostheses. The electrode shift, which may occur during repositioning or donning/doffing of socket, is one main reasons degradation in classification accuracy (CA). acquired from forearm subjects were used pattern recognition-based this study. Multiclass common spatial patterns (CSP) two types schemes, namely versus (CSP-OvO) rest (CSP-OvR), feature extraction improve robustness against shift control. Shift transversal (ST1 ST2) longitudinal (SL1 SL2) direction muscle fibers taken into consideration. We tested nine intact-limb eleven hand wrist motions. CSP features (CSP-OvO CSP-OvR) three commonly features, time-domain (TD) autoregressive (TDAR) variogram (Variog) features. Compared TD significantly improved CA over 10 % all configurations (ST1, ST2, SL1 SL2). TDAR a. CSP-OvO average 5 configurations; b. CSP-OvR ST1, SL2. Variog results demonstrated that