EEG (electroencephalography) is a non-invasive technique for measuring electrical activity at the surface of the scalp. The brain produces small electrical potentials as neurons fire collectively, and EEG picks up those potentials through electrodes that sit on the scalp — often arranged in a stretchy fabric cap that holds 32 or 64 of them in fixed positions.
The signals are small, in the microvolt range (typically 10-100 μV at the scalp), and noisy. Consumer-grade headsets like the Muse have only a handful of electrodes; clinical and research systems have many more, with electrode placement following the standardised 10-20 system (named for the 10% and 20% spacings between landmarks like nasion and inion). The signals carry information about cognitive states, sleep stages, and certain neurological conditions, but extracting that information requires preprocessing — filtering, Normalization, Feature extraction — before any classifier sees them.
EEG activity is conventionally split into five frequency bands, each loosely associated with a brain state: delta (0.5-4 Hz, deep sleep), theta (4-8 Hz, drowsiness, meditation), alpha (8-13 Hz, relaxed wakefulness with eyes closed), beta (13-30 Hz, active thinking, focus), and gamma (30-100 Hz, high-level cognitive processing). Band-power features computed from these ranges are the standard inputs to most EEG classifiers.
A multichannel EEG recording is often visualized as a topographic map showing which regions of the scalp are most active at each moment. The temporal resolution is excellent (samples come in at hundreds of Hz) but the spatial resolution is limited by how blurred the signal becomes passing through skull and scalp.
EEG is one of the sensors most often used in engineering and clinical data-science work, alongside the IMU and ECG electrodes.