SSVEP Frequency Detection โ€” Methods Comparison
Classical FFT  ยท  CCA  ยท  TRCA  โ€”  Observe how each method handles noise, epoch length and number of channels
0 dB
2.0 s
8
๐Ÿ“Š FFTClassical Fourier
โ€”

โœ” Pros

  • Simple & computationally fast
  • No training data needed
  • Interpretable PSD visualization

โœ˜ Cons

  • Single channel โ€” no spatial filtering
  • Resolution ฮ”f = 1/T (needs long epochs)
  • Poor robustness at low SNR
  • Ignores harmonic structure
๐Ÿ”— CCACanonical Correlation
โ€”

โœ” Pros

  • Multi-channel โ€” better SNR than FFT
  • Exploits harmonics (f, 2f, 3fโ€ฆ)
  • No training data required
  • Closed-form solution

โœ˜ Cons

  • Reference signals are generic (not subject-specific)
  • No optimized spatial filter per subject
  • Still degrades at low SNR
๐Ÿง  TRCATask-Related Component
โ€”

โœ” Pros

  • Subject-specific spatial filter
  • Maximises inter-trial reproducibility
  • Highest accuracy especially at low SNR
  • Works with short epochs

โœ˜ Cons

  • Requires training trials per subject
  • More complex implementation
  • Sensitive to non-stationarities
  • Needs multiple EEG channels
Detection Accuracy vs SNR โ€” Theoretical Comparison  (4-class ยท 25% chance level ยท typical literature values)