From: Brain–computer interface robotics for hand rehabilitation after stroke: a systematic review
Study | Pre-processing | Feature extraction | Classification | Hand task |
---|---|---|---|---|
Ang et al. [75] | Band-pass (0.05–40 Hz) | Filter Bank Common Spatial Pattern (FBCSP) algorithm [105] | Calibration model (unspecified) | MI vs rest |
Barsotti et al. [76] | Band-pass (8–24 Hz) | ERD (β and µ-decrease), CSP | SVM with linear kernel | MI vs rest |
Bauer et al. [97] | Band-pass (6–16 Hz using zero-phase lag FIR | ERD (β-decrease) | Linear autoregressive model based on Burg Algorithm | MI vs rest |
Bundy et al. [77] | Unspecified | ERD (β and µ-decrease) | Linear autoregressive model | MI (affected, unaffected) vs rest |
Chowdhury et al. [78] | Band-pass (0.1 Hz-100 Hz), Notch (50 Hz) | CSP Covariance-based, ERD/ERS (β and µ-change) | SVM with linear kernel, Covariate Shift Detection (CSD)-based Adaptive Classifier | left vs right MI |
Coffey et al. [92] | Band-pass (0.5 Hz-30 Hz), Notch (50 Hz) | CSP Covariance-based | Linear Discriminant Analysis (LDA) classifier | MI vs rest |
Diab et al. [103] | Unspecified | Time epochs (unspecified) | Artificial Neural Network (ANN)-based Feed Forward Back Propagation | Non-MI open vs closed |
Frolov al. [80] | Band-pass (5–30 Hz), FIR (order 101), IIR notch Chebyshev type I filter (50 Hz) | Time epochs (10 s) | Bayesian-based EEG covariance classifier [106] | MI (affected, unaffected) vs rest |
Ono et al. [81] | Band-pass (0.5–30 Hz), notch (50 or 60 Hz) | Time epochs (700 ms), ERD (µ-decrease) | Linear Discriminant Analysis (LDA) classifier | MI vs rest |
Ramos-Murguialday et al. [95] | Unspecified | Time epochs (5 s), Spatial filter, ERD/ERS (β and µ-change) | Linear autoregressive model | MI vs rest |
Vukelic and Gharabaghi [99] | High-pass (unspecified) | ERD (β-decrease) | Linear autoregressive model based on Burg Algorithm | MI vs rest |
Witkowski et al. [101] | Band-pass (0.4–70 Hz), Laplacian filter | ERD/ERS (β and µ-change) | Linear autoregressive model based on Yule-Walker algorithm | MI vs rest |