1) Electrodes placed on the scalp (e.g. per the 10-20 system) record potential differences produced by neuronal activity. 2) The weak signals (microvolts) are amplified and sampled (analog-to-digital conversion). 3) The signal is filtered and cleaned of artifacts (motion, blinks, EMG, 50/60 Hz mains). 4) Features are extracted (band power, ERP, connectivity) and then interpreted or classified (e.g. by a machine-learning model) for BCI or mental-state analysis.
Enables direct, non-invasive real-time readout of brain activity, providing an objective signal about the user's mental states and intentions that cannot be obtained from conventional input devices.
Scalp sensors (wet or dry) recording the brain's electrical potentials.
Official
Amplifies the weak (microvolt) signals and samples them into digital form.
Official
Filtering, artifact removal and feature extraction (bands, ERP) from the EEG signal.
Official
The EEG signal is weak (microvolts) and prone to motion, muscle and mains artifacts.
EEG localizes brain sources poorly (smearing through the skull).
Hans Berger records the first human electroencephalogram.
Berger publishes "Über das Elektroenkephalogramm des Menschen".
EEG becomes the core signal for passive BCIs and neuroadaptive technology (Zander & Kothe).
Time complexity: O(kanaly × probki) na okno. Space complexity: O(kanaly × bufor).
Low signal-to-noise ratio (microvolts), artifacts (motion, blinks, EMG, mains) and limited spatial resolution are the main limitations, not compute.
Number of electrodes (e.g. 1-8 for wearables, 32-256 for research).
Wet (gel) vs dry.
E.g. 250-1000 Hz.
Continuous, parallel sampling of all channels.
Continuous signal acquisition, no routing.
Channels and features can be processed independently/in parallel.
EEG acquisition and processing are lightweight; they run on CPU/embedded devices. The key is dedicated analog hardware (electrodes + amplifier).
