LIBRARYEEG Frequency Bands & Filtering
EEG isn't one signal; it's overlapping rhythms at different frequencies. The classic bands (delta to gamma), what they're associated with, and why you bandpass-filter.
Raw EEG isn't a single wave; it's many rhythms layered on top of each other, oscillating at different speeds. By long convention those speeds are grouped into named frequency bands, and each band is loosely associated with particular brain states. The associations are tendencies, not labels: a band is a frequency range, not a one-to-one readout of what someone is doing.
EEG · FREQUENCY BANDS
One signal, many rhythms
mu rhythm sits in alpha's ~8–12 Hz band, but over the motor cortex
| Band | Frequency | Loosely associated with |
|---|---|---|
| Delta | ~0.5–4 Hz | Deep (slow-wave) sleep |
| Theta | ~4–8 Hz | Drowsiness, deep relaxation, memory tasks |
| Alpha | ~8–12 Hz | Relaxed wakefulness, eyes closed |
| Beta | ~13–30 Hz | Alert, active concentration |
| Gamma | ~30+ Hz | High-level / integrative processing |
The sensorimotor 'mu' rhythm sits in the same ~8–12 Hz range as classic alpha, but it lives over the motor cortex and reflects motor activity, not eyes-closed relaxation. So a band alone doesn't identify a signal; where on the scalp it appears and what task produced it matter just as much. That spatial-plus-spectral fingerprint is exactly what motor-imagery decoders exploit.
Why you bandpass-filter EEG
Because the useful signal usually lives in one band, the first real signal-processing step is almost always a bandpass filter: keep the frequencies your task cares about (say mu and beta for motor imagery) and discard the rest, including slow drift below and high-frequency muscle noise above. (A 50/60 Hz notch may follow to knock down mains hum, though, as the noise guide explains, that's a last resort, not the main defense.)
▸Deep dive· Go deeper: theta, drowsiness, and why filtering isn't free
Frontal theta tends to rise with drowsiness and sustained cognitive load, a well-studied general EEG phenomenon and one reason fatigue-monitoring is a classic EEG application (mental-fatigue meta-analysis, 2025). But filtering always has a cost: every filter distorts phase and timing somewhat, a steep filter rings, and a notch removes real signal at its frequency. The craft is keeping the band you need while disturbing it as little as possible.
References
Keep going
Reading these bands cleanly starts with a quiet front-end: the build in the OTD Academy EEG front-end project.
One Thousand Drones Academy · reviewed June 2026
Coming soon
8-Channel EEG Front-End on ESP32 →Design the analog board that reads real brainwaves: the BCI.