One or two plain sentences per term, with the module that teaches it. If a word in your notes file is not here, add it to your questions for Alex.
A #
Absolute band power. Power in a frequency band in µV², the area under the spectrum between the band edges. Depends on electrode contact and skull thickness as well as the brain. → 10
Alpha (8-13 Hz). The rhythm of relaxed wakefulness, strongest at the back of the head with eyes closed. The single most reliable thing in human EEG. → 02
Alpha blocking (alpha reactivity). The drop in posterior alpha when the eyes open. In the course sample, O1 alpha is about 18 times stronger with eyes closed. → 06
Amplitude. How big a signal is, in microvolts (µV). Scalp EEG is typically tens of µV. → 01
Annotation. A labelled time span stored with a recording, such as a marked artifact or an event. Sample s10 stores each cut segment as one. → 09
Artifact. Anything in the recording that does not come from the brain: eyes, muscle, heart, mains, electrodes, movement. → 05
Artifact gate. A rule that pauses feedback (or marks data bad) when a window looks like an artifact, for example when its amplitude is too large. → 13
ASR (artifact subspace reconstruction). A method that repairs short high-amplitude bursts by comparing each moment against a clean reference segment and reconstructing the parts that deviate. The standard choice for real-time use. → 09
Average reference. Re-referencing each channel to the mean of all channels. A common default for dense montages. → 08
B #
Band-pass filter. A filter that keeps one range of frequencies, such as 1-40 Hz; a high-pass and a low-pass combined. → 07
Band-power feature. A number summarising the power in one band over a window of signal; the feature most neurofeedback protocols train on. → 10
Beta (13-30 Hz). Active thinking and motor activity. High beta (roughly 20-30 Hz) overlaps with muscle activity. → 02
Bipolar montage. Displaying each channel as the difference between two neighbouring electrodes, which shows local activity and cancels what they share. → 03, 08
Blink. A big, slow deflection largest at Fp1 and Fp2, fading toward the back. The eyeball is an electrical dipole and the lid moving over it changes the field. → 05
C #
Causal filter. A filter that only uses past samples, so it can run in real time. It delays and distorts the signal slightly. → 07, 13
Channel. One recorded signal, usually one electrode against the reference, shown as one row in a raw plot. → 04
Coherence (HRV). How much of the heart-rhythm spectrum is concentrated in one narrow peak; high when the RR curve is smooth and sine-like during paced breathing. One of the three HRV Training modes. → 14
Component (ICA). One of the source signals ICA unmixes from the recording, with a scalp map and a time course. → 09
D #
DC offset. A large constant voltage, often tens of thousands of µV in raw dry-electrode files, from the electrode-skin junction rather than the brain. A high-pass filter removes it. → 07
Delta (0.5-4 Hz). Slow waves of deep sleep. In awake recordings, large delta is often eye movement or drift. → 02
Drift. A slowly wandering baseline from sweat, changing electrode contact or movement. → 05
E #
ECG (electrocardiogram). A recording of the heart's electrical activity; the source of R-peaks for HRV. Heartbeat can also leak into EEG as an artifact. → 14
EDF / EDF+ (European Data Format). The standard file format for EEG and other physiological recordings. EDF+ adds annotations. → 06
Electrode pop. A sudden jump confined to one electrode when its contact changes abruptly. → 05
EMG (electromyogram). Electrical activity from muscles. In EEG, high-frequency fuzz, worst on temporal channels and with jaw clenching. → 05
Epoch. A fixed window of data cut around an event, such as -0.2 to 0.8 s around a stimulus. → 11
ERP (event-related potential). The average response to a repeated event, revealed by averaging many epochs so background EEG cancels out. → 11
Eyes closed / eyes open (EC/EO). The two standard resting conditions. Comparing them is the quickest test that a recording works. → 06
F #
Feedback. The tone, bar or game the trainee perceives when the protocol's conditions are met. → 13
FIR / IIR filter. Two families of digital filter. FIR filters (MNE's default) are stable and can be zero-phase; IIR filters (such as Butterworth) are shorter and common in real time. → 07
fMRI / MEG. Other ways to measure the brain. fMRI has good spatial but poor time resolution; MEG measures magnetic fields with millisecond resolution and less smearing than EEG, but needs a shielded room. → 01
Frequency band. A conventional range of frequencies (delta, theta, alpha, beta, gamma). Edges vary by author by a hertz or two. → 02
G #
Gamma (30+ Hz). Fast activity, hard to separate from muscle at the scalp. → 02
H #
HF (high frequency, HRV). Heart-rhythm power at 0.15-0.4 Hz, the breathing band; dominated by RSA. → 14
High-pass filter. Removes frequencies below the cutoff, such as drift and DC offset with a 0.5-1 Hz high-pass. → 07
HRV (heart rate variability). Beat-to-beat variation in the interval between heartbeats. → 14
I #
ICA (independent component analysis). Unmixes the recording into statistically independent components so blink, eye-movement, heartbeat and muscle components can be removed. → 09
ICLabel. A classifier (in Python, mne-icalabel) that labels ICA components as brain, eye, muscle, heart, line noise or other. → 09
Inhibit band. A band that must stay below its threshold for feedback to be given. → 13
L #
Latency. The delay between a brain event and the feedback it triggers, from filters, window length, smoothing and display. → 13
LF (low frequency, HRV). Heart-rhythm power at 0.04-0.15 Hz. Rises sharply with slow paced breathing near 6 breaths per minute. One of the three HRV Training modes. → 14
LF/HF ratio. LF power divided by HF power. Widely reported, widely criticised, and heavily affected by breathing rate. → 14
Line noise (mains hum). Interference at the mains frequency, 60 Hz in North America and 50 Hz in most other places, and its harmonics. → 05, 07
Linked mastoids. A reference made from the average of electrodes behind both ears. → 08
Low-pass filter. Removes frequencies above the cutoff, such as muscle and high-frequency noise with a 40-45 Hz low-pass. → 07
M #
Mean HR. Average heart rate in beats per minute; 60 000 divided by the mean RR interval in ms. → 14
MNE-Python. The standard open-source EEG/MEG analysis library in Python, used in every lab. → 06
Montage. Either the set of electrode positions (MNE's meaning) or the way channels are combined for display, such as referential or bipolar (the clinical meaning). → 03
N #
Neurofeedback. Training a person to change a feature of their own EEG by giving them real-time feedback when it moves the right way. → 13
Notch filter. Removes a narrow band around one frequency, usually the mains frequency. → 07
Nyquist frequency. Half the sampling rate; the highest frequency a recording can represent. 125 Hz for a 250 Hz recording. → 07
O #
Operant conditioning. Learning driven by reward: behaviour (here, a brain state) that is followed by reward becomes more likely. → 13
P #
Post-synaptic potential. The voltage change in a neuron when it receives input. Summed across millions of aligned pyramidal neurons, these produce scalp EEG. → 01
PPG (photoplethysmography). An optical sensor measuring the blood-volume pulse with each heartbeat; used by most wearables for HRV. → 14
Protocol. The full specification of a neurofeedback exercise: site, reward band, inhibit bands and thresholds. → 13
PSD (power spectral density). How a signal's power is distributed across frequencies; "the spectrum". → 10
Pyramidal neurons. The large cortical neurons whose aligned dendrites make their summed activity visible at the scalp. → 01
R #
R-peak. The tall sharp spike of each heartbeat in the ECG, used to time beats precisely. → 14
Raw trace. The time-domain plot of a recording: time across, one row per channel, a µV scale bar. → 04
Re-referencing. Changing what every channel is measured against after recording, such as switching to an average reference. → 08
Reference electrode. The electrode every channel's voltage is measured against. Choosing it changes what you see. → 03, 08
Rejection. Removing bad channels or bad time segments rather than repairing them. → 09
Relative band power. Band power divided by total power over a broad range; removes overall scale differences between people and sessions. → 10
Resonance frequency breathing. Breathing at the rate (about 6 per minute, 0.1 Hz) where breathing and the baroreflex reinforce each other and heart-rate swings are largest. → 14
Reward band. The band that must be above its threshold for feedback to be given. → 13
RMSSD. Root mean square of successive differences between RR intervals; a short-window index of vagal activity. → 14
RR interval (NN interval). The time between successive heartbeats, in ms. → 14
RSA (respiratory sinus arrhythmia). Heart rate rising on the in-breath and falling on the out-breath. RSA amplitude is one of the three HRV Training modes. → 14
S #
Sampling rate. How many samples per second the amplifier records, in Hz. The course samples are 242-251 Hz; always read it from the file. → 04
SDNN. Standard deviation of NN intervals; overall heart-rate variability. Depends on recording length. → 14
SMR (sensorimotor rhythm). Roughly 12-15 Hz activity over the sensorimotor strip (C3, Cz, C4); a classic neurofeedback reward band. → 13
T #
10-20 system. The international scheme for naming and placing scalp electrodes at 10% and 20% distances between skull landmarks. Odd numbers left, even right, z midline. → 03
Theta (4-8 Hz). Drowsiness, memory, and frontal-midline theta during mental effort. Blinks and eye movements also land here. → 02
Theta/beta ratio. Theta power divided by beta power, typically at frontal or central sites; a common neurofeedback and qEEG feature. → 10
Threshold. The level a band's power is compared against to decide whether to give feedback. Usually adaptive. → 13
Topomap. A head map colouring each point by a value such as alpha power, interpolated between electrodes. → 06
V #
Vagal activity. Parasympathetic activity carried by the vagus nerve, which slows the heart and drives RSA. → 14
Volume conduction. The spread of electrical activity through brain, skull and scalp, which smears each source across many electrodes and limits spatial resolution. → 01
W #
Welch's method. Estimating a spectrum by cutting the signal into overlapping windows, tapering and transforming each, and averaging. → 10
Window (analysis window). The stretch of signal used to compute one feature value, such as the last 1 s in a neurofeedback loop. → 10, 13
Z #
Zero-phase filter. A filter applied forwards and backwards so it does not shift the signal in time; needs the whole recording, so it is offline only. → 07