Introduction
Materials and Methods
Motor Imagery BCI System
EMG-Based Neuromuscular Interface
Results
Motor Imagery BCI System
Dataset | Training data size | Decoding accuracy (all components) | Decoding accuracy (choice) |
---|---|---|---|
A1 | 360 s | 0.66 | 0.74 |
P1 | 270 s | 0.46 | 0.51 |
P2 | 270 s | 0.32 | 0.40 |
K1 | 270 s | 0.75 | 0.78 |
K2 | 270 s | 0.55 | 0.63 |
Synchronous/asynchronous | Num. of channels | Num. of commands | Training duration | Online feedback | Accuracy (%) | Source | Methods |
---|---|---|---|---|---|---|---|
Synchronous | 64 | 4 | 6–9 sessions, 240–360 trials total | no | 63 | [22] | AAR, SVM |
Synchronous | 64 | 4 | 6 sessions, 192 trials each | yes | 72 | [1] | Bandpass filtering, CSP, regularized LDA |
Asynchronous | 2 | 2 | 4 months of usage of Graz-BCI | n/a | >90 | [10] | Decision rule based on power threshold in frequency band |
Asynchronous | 5 | 3 | A couple of sessions, 3 min (30 trials) each | yes | 75 | [16] | Local neural classifier [9] |
Asynchronous | 8 | 3 | ~4 sessions, 5 min each | yes | ~70 | [27] | wavelets, CSFP |
Asynchronous | 32 | 4 | ~6 min | no | 81 | Designed BCI | Bandpass filtering, CSP, LDA, MLP |
EMG-Based Neuromuscular Interface
Indicator measured | SRD-1 | [4] | [25] | [8] | [5] | [6] | [3] |
---|---|---|---|---|---|---|---|
Recognition accuracy | 98.3% | – | 96% | >90% | – | ~90% | – |
Control | Command and proportional control | Consistent proportional control | Motion pattern recognition. Proportional control | Proportional control | Proportional control | Command and proportional control | Motion pattern recognition. Proportional control |
Classifier | Hybrid neural network | Linear discriminant analysis | Linear discriminant analysis | Artificial neural network (perceptron) | Artificial neural network (perceptron) | Linear regression | Linear discriminant analysis |
Number of gestures/commands | 9 gestures | 5 gestures | 2 DoF, 5 gestures | 3 DoF | 2 DoF, 4 gestures | 2 DoF, 4 gestures | 8 gestures |
Number of EMG channels/sensors | 8 for recording + 1 reference | 5 | 6 | 7 pairs for each forearm | 192-channel electrode array in the monopolar configuration | 4 for each type of electrode | 12 pairs of bipolar electrodes |