SimNeXt-EEG: Compact Motor Imagery EEG Decoding via Parameter-Free Temporal Attention and Separable Refinement
Published, Early Access
Research approach
Built a compact motor-imagery EEG decoder using multiscale convolution, parameter-free temporal attention and separable refinement.
Results
BCI IV 2a session-independent accuracy: 70.81 ± 0.49%; 1.84K parameters; 1.46 ms desktop CPU latency at batch size 1.
| Stage | Operation / output |
|---|---|
| MSTF | Temporal kernels 32, 80 / 8 channels |
| STE | Spatial separable convolution + SimAM |
| STR | Temporal separable convolution + SimAM |
| Classification | Flatten, Linear / 4 classes |
| Model | Parameters | CPU latency |
|---|---|---|
| EEGNet | 3.44K | 0.79 ms |
| MSVTNet | 75.49K | 3.78 ms |
| SimNeXt | 1.84K | 1.46 ms |
InterpretationBCI Competition IV 2a, session-independent accuracy: 70.81 ± 0.49%. Complexity: 1.84K parameters and 20.69 MFLOPs. Latency: 1.46 ms on a desktop CPU, batch size 1; not a wearable-device measurement.
All publications