Dictionary

Every metric, signal-processing concept, machine-learning term, and NeuroSkill™-specific phrase — defined in plain language. Metric references link to their primary peer-reviewed source; see the full reference list for details.

#

10-20 System EEG Standard: The international standard for scalp electrode placement. Positions are named by lobe (F=frontal, T=temporal, P=parietal, O=occipital, C=central) and hemisphere (odd=left, even=right, z=midline). Muse uses AF7, AF8 (anterior frontal) and TP9, TP10 (temporal-parietal) per this system.

A

AASM AASM Standard: American Academy of Sleep Medicine — the body that publishes official clinical guidelines for polysomnography and sleep staging. NeuroSkill™'s automatic sleep staging algorithm follows AASM rules for scoring Wake, N1, N2, N3 (deep/slow-wave), and REM using EEG band-power ratios on 30-second epochs.
Alpha α Metric · Band Powers: Σ PSD(8–13 Hz); The dominant rhythm during relaxed wakefulness with eyes closed. Alpha power decreases ('desynchronises') when you open your eyes, concentrate, or become anxious. Posterior alpha is strongest at occipital sites; frontal alpha is key to FAA.; Klimesch (1999); Nunez & Srinivasan
Alpha Coherence Metric · Connectivity & Asymmetry: cross-spectral coherence in α band; Measures how synchronised alpha oscillations are between frontal electrode pairs. High coherence suggests strong functional connectivity. Changes with cognitive task, meditation, and pathology. Computed via the cross-spectral density.; Lachaux et al. (1999)
Alpha Peak Frequency APF Metric · Spectral Shape: argmax PSD in 8–13 Hz; The frequency at which alpha power is maximal, typically 9–11 Hz in healthy adults. APF slows with age and cognitive decline, and speeds up with caffeine or cognitive training. A fundamental individual trait — stable within a person day-to-day.; Klimesch (1999)
Artifact Signal Processing: Non-neural signal contamination in EEG. Common sources: eye blinks and saccades (EOG artifact in frontal channels), jaw clenches and facial muscles (EMG in temporal channels), electrode cable motion, power-line interference (50/60 Hz), and sweat potentials. NeuroSkill™ flags per-channel artifact levels and excludes high-artifact epochs from sleep staging.

B

Baevsky Stress Index SI Metric · PPG / HRV: AMo / (2 × Mo × MxDMn); A Russian aerospace-derived metric: the Amplitude of the Mode (AMo) divided by twice the product of Mode (Mo) and range (MxDMn) of the IBI histogram. Higher SI = greater sympathetic dominance / stress. Used in space medicine to assess astronaut autonomic status.; Baevsky et al. (1984)
Band-Power Slope BPS Metric · Spectral Shape: log-log fit slope of PSD; The 1/f aperiodic exponent of the power spectrum. Steeper slopes (more negative) indicate less high-frequency activity. The 1/f slope has been linked to excitation-inhibition balance in neural circuits — flatter slopes suggest more excitation.; Donoghue et al. (2020)
Bandpass Filter Signal Processing: A filter that attenuates frequencies outside a defined range. NeuroSkill™ applies a 0.5–50 Hz bandpass to each EEG channel before epoch analysis — removing DC drift below 0.5 Hz and high-frequency noise above 50 Hz while preserving the physiologically relevant EEG spectrum.
Beta β Metric · Band Powers: Σ PSD(13–30 Hz); Linked to active thinking, focus, alertness, and motor planning. High beta can indicate anxiety or stress ('beta buzz'). Low beta relative to theta is characteristic of inattentive states. Used in the TBR ratio for ADHD assessment.; Pope et al. (1995); Monastra et al. (1999)
Beta/Alpha Ratio BAR Metric · Cross-Band Ratios: β / α; Higher BAR suggests greater cortical activation or stress. A useful marker for arousal level — increases during demanding tasks, anxiety, or when alpha suppresses. Low BAR = relaxed; high BAR = activated/stressed.; Angelidis et al. (2016)
Blink Count Metric · Events & IMU: cumulative frontal δ spikes; Total eye blinks detected from frontal electrode artifacts (AF7/AF8). Each blink produces a characteristic large-amplitude slow-wave deflection. Blink rate (15–20/min typical) increases with fatigue, dry eyes, and cognitive load.; Maddox et al. (2003)
Blink Rate Metric · Events & IMU: blinks / time; Blinks per minute — a spontaneous behavioral marker. Elevated blink rate correlates with dopamine activity, fatigue, and information processing demands. Drops during focused visual tasks (e.g. reading).; Maddox et al. (2003)
Bluetooth LE BLE Protocol: Bluetooth Low Energy — the wireless protocol used by Muse headsets. Muse streams 4-channel EEG at 256 Hz, PPG, and 9-axis IMU data over BLE GATT characteristics. NeuroSkill™ manages the BLE connection via the host OS Bluetooth stack (Core Bluetooth on macOS, BlueZ on Linux).
Brain-Computer Interface BCI Neuroscience: A system that establishes a direct communication pathway between the brain and an external device, using recorded neural signals as control inputs. NeuroSkill™ supports BCI research workflows via its WebSocket API, labeled data export, and motor-gesture detection (blinks, jaw clenches, head nods/shakes) as voluntary control signals.

C

Calibration NeuroSkill™: A structured recording session in NeuroSkill™ where you alternate between defined mental actions (e.g. relax / focus) for fixed durations, with configurable breaks. Calibration establishes your personal metric baseline and creates labeled embeddings used for personalised similarity search and state classification.
Cognitive Load Metric · Brain Scores: θ × β / α²; Increases with mental effort and working memory demand. Frontal theta rises with task difficulty while alpha suppresses. High cognitive load over extended periods may indicate mental fatigue risk.; Borghini et al. (2014)
Cosine Distance Machine Learning: 1 − (A·B) / (‖A‖ · ‖B‖); A similarity measure between two vectors defined as 1 − cosine similarity. Independent of vector magnitude — only direction matters. Cosine distance = 0 means the vectors point in the same direction (identical brain states); = 1 means orthogonal (completely dissimilar). Used by NeuroSkill™'s HNSW index to rank search results.

D

Delta δ Metric · Band Powers: Σ PSD(0.5–4 Hz); The slowest oscillation, dominant during deep sleep (N3). High delta in waking states may indicate drowsiness, brain injury, or subcortical processing. In healthy awake adults, delta is typically low relative to alpha and beta.; Knyazev (2012); Nunez & Srinivasan
Delta/Theta Ratio DTR Metric · Cross-Band Ratios: δ / θ; Higher DTR indicates deeper states — prominent during deep sleep (N3). In waking, very high DTR may flag abnormal slow-wave activity. Useful for distinguishing sleep stages and monitoring sedation depth.; Knyazev (2012)
DFA Exponent DFA α Metric · Complexity: log-log slope of F(n); Detrended Fluctuation Analysis quantifies long-range temporal correlations. α = 0.5: uncorrelated (white noise). α = 1.0: 1/f noise (pink noise, healthy EEG). α = 1.5: Brownian motion. Healthy EEG typically sits near 0.8–1.0; deviations may indicate pathology.; Peng et al. (1995)
Drowsiness Metric · Brain Scores: θ / β increase pattern; Tracks the progressive shift from alert (low theta/beta ratio) to drowsy (high theta, dropping beta). Used in fatigue research for pilots, drivers, and shift workers.; Lal & Craig (2002)

E

Electroencephalography EEG Neuroscience: The recording of electrical potentials produced by synchronised activity of large populations of neurons, sampled at the scalp via electrodes. Amplitudes are on the order of microvolts (µV). NeuroSkill™ processes 4-channel EEG from Muse (AF7, AF8, TP9, TP10) at 256 Hz for all spectral, connectivity, and complexity metrics.
Electromyography EMG Signal Processing: Recording of electrical activity from muscle fibres. In EEG, EMG is a dominant artifact source — jaw clenches produce characteristic broadband (>100 Hz) bursts at temporal electrodes (TP9, TP10), while facial muscles contaminate frontal channels. NeuroSkill™ uses temporal EMG bursts to detect and count jaw-clench events.
Electrooculography EOG Signal Processing: Recording of electrical potentials caused by eye movements and blinks. The corneoretinal potential creates a large-amplitude slow-wave artifact in frontal EEG channels (AF7, AF8) whenever the eyelids move. NeuroSkill™ uses these characteristic frontal artifacts to detect and count blink events.
Embedding Machine Learning: A dense, fixed-length vector representation of a data point in a continuous metric space. In NeuroSkill™, each 5-second EEG epoch is encoded by the ZUNA neural network to a 32-dimensional floating-point vector. Semantically similar brain states map to nearby points in this embedding space, enabling vector search.
Engagement Metric · Brain Scores: β / (α + θ) (softer curve); Similar to Focus but with a gentler sigmoid — tracks sustained engagement rather than peak concentration. Useful for monitoring long tasks where focus fluctuates but overall engagement remains.; Pope et al. (1995)
Epoch Signal Processing: A fixed-length, time-windowed segment of continuous EEG used as the unit of analysis. NeuroSkill™ uses 5-second epochs (1,280 samples at 256 Hz) with configurable overlap (0–4.5 s). Each epoch produces one complete update of all 60+ metrics, one ZUNA embedding, and one HNSW index entry.

F

Fast Fourier Transform FFT Signal Processing: X[k] = Σ x[n] · e^(−2πi·nk/N); An efficient algorithm (O(N log N)) for computing the Discrete Fourier Transform, decomposing a time-domain signal into its component frequencies. NeuroSkill™ applies a 512-sample Hann-windowed FFT to each channel in each epoch to produce the power spectral density (PSD) that feeds all spectral metrics.
Focus Metric · Brain Scores: β / (α + θ); Measures active directed attention. High when beta dominates (concentrated work, problem-solving). Low during mind-wandering or relaxation. The ratio captures the brain's shift from idle alpha/theta to task-engaged beta.; Pope et al. (1995)
Frontal Alpha Asymmetry FAA Metric · Connectivity & Asymmetry: ln(AF8 α) − ln(AF7 α); The most-studied EEG asymmetry metric. Positive FAA (more right-frontal alpha = less right activity) suggests approach motivation, positive affect, engagement. Negative FAA suggests withdrawal, avoidance, or anxiety. Validated specifically on Muse by Cannard et al. (2021).; Coan & Allen (2004); Cannard et al. (2021)

G

Gamma γ Metric · Band Powers: Σ PSD(30–50 Hz); The fastest measured oscillation, associated with cross-modal binding, perception, and higher cognitive processing. Gamma is often noisy in consumer EEG due to muscle artifact (EMG), so NeuroSkill™ displays it but flags the caveat.; Canolty et al. (2006)

H

Hann Window Signal Processing: w(n) = 0.5 · (1 − cos(2πn / (N−1))); A bell-shaped tapering function applied to each FFT segment before transform, reducing spectral leakage. Without windowing, power from one frequency would 'leak' into adjacent bins due to the discontinuity at segment edges. The Hann window smoothly tapers signal amplitude to zero at both ends.
Head Pitch Metric · Events & IMU: complementary filter on accel + gyro; Forward/backward head tilt from the Muse IMU. Pitch increases (looking down) during drowsiness — 'head drops' are a classic fatigue indicator. Monitored continuously.; Madgwick et al. (2011)
Head Roll Metric · Events & IMU: complementary filter on accel + gyro; Side-to-side head tilt. Large roll deviations may indicate sleep position changes or vestibular responses. Combined with pitch for full head orientation tracking.; Madgwick et al. (2011)
Headache Index Metric · Headache & Migraine Correlates: 0.35·β↑ + 0.35·α↓ + 0.30·BAR↑; Cortical hyperexcitability correlate: elevated relative beta, suppressed alpha, and a high Beta/Alpha Ratio (BAR) — patterns observed interictally in headache research. Higher values suggest greater cortical excitability. Research indicator only; not a clinical diagnostic.; Bjørk et al. (2009)
Heart Rate HR Metric · PPG / HRV: 60 / mean(IBI); Beats per minute from IR peak detection. Muse PPG uses the forehead vasculature — slightly different from wrist or fingertip but validated for trend monitoring. Typical resting range: 60–100 bpm.; Allen (2007)
Heart Rate Variability HRV Physiology: The natural, beat-to-beat variation in the interval between heartbeats (IBI). Not to be confused with arrhythmia — HRV is the healthy oscillation driven by breathing, autonomic nervous system activity, and circadian rhythms. Higher HRV generally reflects greater autonomic flexibility and cardiovascular health. NeuroSkill™ computes RMSSD, SDNN, pNN50, and LF/HF from the IBI sequence.
Higuchi Fractal Dimension HFD Metric · Complexity: log-log slope of L(k); Quantifies the fractal structure ('roughness') of the time series. HFD = 1 for a smooth curve; HFD → 2 for increasingly complex/jagged signals. Decreases during sleep and anaesthesia. More discriminative than spectral features for some pathologies.; Higuchi (1988)
Hjorth Activity Metric · Complexity: var(x); Signal variance — represents total power in the time domain. Equivalent to the area under the power spectrum. Increases with signal amplitude. The simplest complexity measure.; Hjorth (1970)
Hjorth Complexity Metric · Complexity: mobility(x') / mobility(x); Measures how much the signal deviates from a pure sine wave. Complexity = 1 for a perfect sine. Higher values indicate more bandwidth / irregular shape. Sensitive to artifacts and pathological EEG patterns.; Hjorth (1970)
Hjorth Mobility Metric · Complexity: √(var(x') / var(x)); Estimates the mean frequency of the signal. Higher mobility = more high-frequency content. A computationally cheap alternative to spectral centroid. Very fast to compute — often used in real-time BCI.; Hjorth (1970)
HNSW Index HNSW Machine Learning: Multi-layer navigable graph with log(N) expected search time; Hierarchical Navigable Small World — a graph-based approximate nearest-neighbour (ANN) data structure. NeuroSkill™ stores all session embeddings in an HNSW file (M=16, ef_construction=200). Queries return the k closest historical brain-state moments by cosine distance in sub-millisecond time, enabling efficient similarity search over thousands of epochs.
Hypnogram Physiology: A time-series chart of sleep stage transitions, with stage on the Y-axis (Wake at top, N3 at bottom, REM typically near N2) and time on the X-axis. Also called a 'sleep architecture' chart. NeuroSkill™ automatically generates hypnograms from sessions ≥30 minutes, with per-stage duration, efficiency, and onset latency statistics.

I

Impedance EEG: The opposition to current flow between the electrode and scalp tissue, measured in ohms (Ω). Lower impedance = better electrical contact = higher signal quality. Wet-gel electrodes typically achieve <5 kΩ; dry electrodes (like Muse) rely on pressure contact and achieve ~50–200 kΩ. NeuroSkill™ continuously estimates impedance and displays a per-channel quality indicator.
IMU IMU Hardware: Inertial Measurement Unit — a sensor combining a 3-axis accelerometer, 3-axis gyroscope, and 3-axis magnetometer (9 degrees of freedom). The Muse IMU provides continuous head orientation and motion data at ~50 Hz. NeuroSkill™ derives head pitch, head roll, stillness, nod count, and shake count from the IMU stream.
Information Integration Φ proxy Metric · Consciousness Metrics: 0.40·coherence + 0.35·PAC + 0.25·PSE → 0–100; Proxy for global information integration (global workspace theory): inter-channel alpha coherence × theta-gamma phase-amplitude coupling × power spectral entropy. Higher values suggest more integrated, brain-wide activity — a signature of conscious states in Integrated Information Theory (IIT) and Global Workspace frameworks.; Tononi (2004)
Inter-Beat Interval IBI Physiology: IBI_n = t_n+1 − t_n (ms); The time elapsed between successive heartbeat peaks in the PPG or ECG waveform, measured in milliseconds. Normal range: 600–1000 ms (60–100 bpm). IBIs are the raw input to all HRV metrics — RMSSD, SDNN, pNN50, and LF/HF are all computed from the sequence of successive IBI values.

J

Jaw Clench Count Metric · Events & IMU: cumulative temporal EMG bursts; Jaw clenches detected from high-frequency EMG artifact at temporal electrodes (TP9/TP10). Voluntary clenches can serve as BCI control signals. Involuntary clenching may indicate stress or bruxism.; Temporal EMG detection
Jaw Clench Rate Metric · Events & IMU: clenches / time; Clenches per minute — baseline should be near zero for relaxed subjects. Elevated rates during sleep may indicate bruxism. Useful as a voluntary control signal in motor-imagery BCIs.; Temporal EMG detection

L

Lab Streaming Layer LSL Protocol: An open-source framework for real-time exchange of time-series data in neuroscience research. LSL is used by OpenBCI GUI and many research EEG systems. NeuroSkill™ does not output LSL natively, but data can be forwarded via the WebSocket API to a bridge script.
Laterality Index Metric · Connectivity & Asymmetry: generalised L/R power asymmetry; Extends FAA to all frequency bands — a comprehensive left-right hemispheric power balance. Positive = more right-dominant; negative = more left-dominant. Useful for tracking hemispheric shifts during different cognitive tasks.; Harmon-Jones & Gable (2010)
Lempel-Ziv Complexity (LZC) LZC Metric · Consciousness Metrics: 0.60·PE + 0.40·HFD → 0–100; Signal information richness, approximated via Permutation Entropy (ordinal complexity) and Higuchi Fractal Dimension (fractal complexity). Higher LZC indicates more complex, information-rich EEG — a signature of conscious states. Inspired by the Perturbational Complexity Index (PCI) used in TMS-EEG consciousness research.; Casali et al. (2013)
LF/HF Ratio Metric · PPG / HRV: power(0.04–0.15 Hz) / power(0.15–0.4 Hz); Low-frequency to high-frequency HRV power ratio. Traditionally interpreted as sympathovagal balance (higher = more sympathetic), though this interpretation is debated. LF includes both sympathetic and parasympathetic components.; Task Force ESC/NASPE (1996)

M

mDNS mDNS Protocol: Multicast DNS — a zero-configuration protocol for service discovery on local networks (RFC 6762). NeuroSkill™ advertises itself as `_skill._tcp.local.` via mDNS so CLI tools and custom clients can automatically discover and connect without needing to know the host IP address or configure port numbers manually.
Meditation Metric · Brain Scores: α / θ elevation pattern; Captures the alpha/theta crossover characteristic of experienced meditators — elevated alpha with gradually rising theta, without beta agitation. Higher scores indicate deeper meditative states.; Lomas et al. (2015)
Migraine Index Metric · Headache & Migraine Correlates: 0.40·δ↑ + 0.35·α↓ + 0.25·laterality; Cortical spreading depression proxy: elevated delta power, alpha suppression, and hemispheric lateralisation — patterns observed interictally in migraine research. Uses the same validated interictal QEEG source as the Headache Index. Research indicator only; not a clinical diagnostic.; Bjørk et al. (2009)
Mood Index Metric · Brain Scores: FAA + α/β pattern; A composite combining frontal alpha asymmetry with overall alpha/beta balance. Higher values suggest more positive, approach-oriented states; lower values suggest withdrawal or negative affect. Exploratory — not a clinical mood measure.; Harmon-Jones & Gable (2010)
Mu Suppression Metric · Connectivity & Asymmetry: α power decrease over sensorimotor regions; Mu rhythms (8–13 Hz over motor cortex) suppress during movement execution and observation. With Muse's temporal electrodes (TP9/TP10), this captures a proxy of mu suppression — useful in motor imagery BCI research.; Pfurtscheller & Lopes da Silva (1999)
Muse 2 MU-02 Hardware: InteraXon Muse 2 headset. 4-channel EEG at AF7, AF8, TP9, TP10 (256 Hz), forehead PPG sensor (IR + red LEDs), 9-axis IMU. Connects via Bluetooth LE. Primary development target for NeuroSkill™. Required for PPG/HRV metrics. The most widely used device with NeuroSkill™.
Muse S MU-03 / MU-04 Hardware: Sleep-focused Muse headset with soft, flexible form factor designed for overnight wear. Same EEG and PPG sensing as Muse 2. Gen 2 (MU-04) adds USB-C charging. Both generations fully supported by NeuroSkill™. Preferred for sleep staging experiments due to improved comfort.

N

Nod Count Metric · Events & IMU: pitch oscillation detection; Head nods detected from pitch oscillation patterns — rapid downward + upward pitch changes. Can be used as a BCI control gesture ('yes' nod) or tracked during microsleep episodes.; IMU processing
Notch Filter Signal Processing: A narrow band-stop filter that removes power-line interference at exactly 50 Hz (European/international) or 60 Hz (North American) from the raw EEG signal. Configurable in NeuroSkill™ → Settings → Signal Processing. Crucial for clean spectral analysis — power-line noise would otherwise contaminate the gamma band.

O

OSC OSC Protocol: Open Sound Control — a UDP-based message protocol for streaming data over networks, used by Muse Direct to stream raw EEG. NeuroSkill™ does not use OSC; it connects directly to Muse hardware via BLE for lower latency and richer access to the full PPG + IMU data stream.

P

Perfusion Index PI Metric · PPG / HRV: AC/DC ratio of PPG; The ratio of pulsatile to non-pulsatile blood flow at the sensor site. Higher PI = stronger peripheral blood flow. Affected by temperature, stress, and vasoconstriction. Typical: 0.5–5%.; Allen (2007)
Permutation Entropy PE Metric · Complexity: -Σ p(π)·log(p(π)); Measures the complexity of ordinal patterns in the signal. PE → 0: very regular (e.g. seizure). PE → 1: completely random. Robust to noise and amplitude changes, making it ideal for EEG where absolute amplitudes vary. Order parameter typically m=3.; Bandt & Pompe (2002)
Phase-Amplitude Coupling PAC Metric · Connectivity & Asymmetry: θ phase × γ amplitude; Measures how strongly gamma amplitude is modulated by theta phase. Strong PAC is associated with successful memory encoding and retrieval — gamma 'bursts' ride on theta waves during learning. Weaker PAC is observed in Alzheimer's and schizophrenia research.; Canolty et al. (2006)
Photoplethysmography PPG Hardware: An optical technique for measuring blood volume changes using infrared and red LED illumination. The Muse 2 and Muse S include a forehead PPG sensor. NeuroSkill™'s PPG pipeline extracts heart rate, RMSSD, SDNN, pNN50, LF/HF, respiratory rate, SpO₂ estimate, perfusion index, and Baevsky Stress Index from the raw PPG waveform.
pNN50 Metric · PPG / HRV: % of |ΔIBIs| > 50 ms; Percentage of successive intervals differing by more than 50 ms. Another parasympathetic marker — higher pNN50 = greater vagal modulation. Typical resting: 5–25%. Often correlated with RMSSD.; Task Force ESC/NASPE (1996)
Power Spectral Density PSD Signal Processing: PSD(f) = |X(f)|² / (N · fs); The distribution of signal power across frequencies (µV²/Hz). Computed via the Welch method (4 overlapping Hann-windowed FFTs averaged together) for each 5-second epoch. The PSD is the fundamental input from which all spectral EEG metrics — band powers, ratios, entropy, centroid, and slope — are derived.
Power Spectral Entropy PSE Metric · Spectral Shape: -Σ p·log(p); Shannon entropy of the normalised power spectrum. PSE = 0 means all power is in one frequency (a pure sine wave). PSE → 1 means power is uniformly distributed (white noise). Typical awake EEG: 0.5–0.8. PSE drops during sleep as delta dominates.; Inouye et al. (1991)

R

Reference Electrode EEG: A nominally electrically inactive electrode used as the voltage reference against which all active channels are measured. Muse uses a linked-mastoid reference (TP9 and TP10 averaged). Reference choice affects the spatial distribution of recorded potentials — particularly relevant for asymmetry metrics like FAA.
Relaxation Metric · Brain Scores: α / (β + θ); Measures calm, wakeful rest. Peaks during eyes-closed relaxation, guided breathing, or meditation when alpha is strong and beta/theta are subdued. Decreasing relaxation during a task often means increasing cognitive demand.; Pope et al. (1995)
Respiratory Rate Metric · PPG / HRV: PPG envelope modulation; Estimated from the respiratory sinus arrhythmia in the PPG waveform. Breathing modulates heart rate — NeuroSkill™ extracts this modulation frequency. Typical: 12–20 breaths/min. Less accurate during irregular breathing.; Charlton et al. (2016)
RMSSD Metric · PPG / HRV: √(mean(ΔIBI²)); Root Mean Square of Successive Differences in inter-beat intervals. The gold-standard time-domain HRV metric, primarily reflecting parasympathetic (vagal) tone. Higher RMSSD = more resilient autonomic nervous system. Typical resting: 20–70 ms.; Task Force ESC/NASPE (1996)

S

Sample Entropy SampEn Metric · Complexity: -ln(A/B); Measures signal regularity — lower values mean more predictable (regular) signals, higher values mean more complex/unpredictable. Unlike approximate entropy, it's less biased for short data. Decreases during epileptic seizures and deep anaesthesia.; Richman & Moorman (2000)
SDNN Metric · PPG / HRV: std(IBI); Standard Deviation of Normal-to-Normal intervals. Reflects total autonomic variability (both sympathetic and parasympathetic). Higher SDNN = greater overall HRV. Highly dependent on recording length — compare only equal-duration sessions.; Task Force ESC/NASPE (1996)
Session NeuroSkill™: A continuous NeuroSkill™ recording from headset connection through disconnection or manual stop. Each session is stored as a SQLite database (metrics, embeddings, labels, sleep stages) plus a raw EEG CSV file at 256 Hz. Sessions are individually timestamped and can be selected for side-by-side comparison in the Compare window.
Shake Count Metric · Events & IMU: roll oscillation detection; Head shakes detected from roll oscillation patterns — rapid left-right-left roll changes. Can serve as a 'no' gesture control in BCI applications.; IMU processing
Signal Quality NeuroSkill™: A per-channel contact quality indicator recomputed every 2.5-second epoch from rolling RMS windows on each EEG channel. Four states: Good (≥95% usable epochs, green), Fair (70–95%, yellow), Poor (40–70%, orange), Off (<40%, red). Sleep staging, HNSW indexing, and some connectivity metrics require all four channels to be Good.
Signal-to-Noise Ratio SNR Metric · Spectral Shape: signal power / noise power; Broadband EEG signal power relative to estimated noise floor. Higher SNR means cleaner data. Low SNR may indicate poor electrode contact, excessive muscle artifact, or environmental interference. Useful for data quality monitoring.; Cohen (2014)
Sleep Staging NeuroSkill™: Automatic classification of 30-second EEG epochs into one of five AASM sleep stages: Wake, N1 (light sleep), N2 (stable sleep), N3 (deep/slow-wave), or REM. Uses delta/theta/alpha/beta band-power ratio heuristics per AASM guidelines. Available for sessions ≥30 minutes. Output: staircase hypnogram with per-stage duration and efficiency statistics.
Spectral Centroid Metric · Spectral Shape: Σ(f × PSD(f)) / Σ PSD(f); The power-weighted average frequency — like the 'centre of mass' of the spectrum. Higher centroid = more high-frequency activity (alertness). Lower centroid = more slow-wave dominance (drowsiness, sleep). A single number that summarises overall spectral balance.; Cohen (2014)
Spectral Edge Frequency 95% SEF95 Metric · Spectral Shape: freq at 95% cumulative power; The frequency below which 95% of total spectral power lies. Used extensively in anaesthesia monitoring — SEF95 drops as sedation deepens. In awake states, typically 20–30 Hz. Values below 10 Hz suggest deep sleep or heavy sedation.; Rampil & Sasse (1980)
SpO₂ Estimate Metric · PPG / HRV: red/IR ratio → calibration curve; Blood oxygen saturation estimated from the ratio of red to infrared PPG. This is an uncalibrated estimate — Muse is not a medical pulse oximeter. Useful for relative trends only. Healthy range: 95–100%.; Allen (2007)
Stillness Metric · Events & IMU: 1 − norm(gyro); Inverse of angular velocity magnitude — how still the head is. Stillness → 1 during quiet meditation or sleep. Stillness → 0 during active movement. A simple but effective behavioral state indicator.; IMU processing
Sympathovagal Balance Physiology: The relative activity of the sympathetic ('fight-or-flight') vs. parasympathetic ('rest-and-digest') branches of the autonomic nervous system. Estimated from HRV: high LF/HF ratio suggests sympathetic dominance (stress, arousal); high RMSSD and pNN50 suggest parasympathetic dominance (calm, recovery). Note: the LF/HF interpretation is debated in current literature.

T

Tauri Framework: A cross-platform desktop application framework using Rust for the system backend and web technologies (HTML/CSS/JS via Svelte) for the UI. NeuroSkill™ targets Tauri 2, enabling macOS, Linux, and Windows builds from a single codebase. Tauri provides OS integration (system tray, notifications, file system, Bluetooth via plugins) with a smaller binary footprint than Electron.
Theta θ Metric · Band Powers: Σ PSD(4–8 Hz); Associated with drowsiness, light sleep (N1), meditation, and memory encoding. Frontal midline theta increases during working memory tasks. Elevated theta with low beta can indicate reduced cortical arousal (relevant to ADHD research).; Klimesch (1999); Putman et al. (2010)
Theta/Alpha Ratio TAR Metric · Cross-Band Ratios: θ / α; Higher TAR indicates reduced alertness. Increases during drowsiness, fatigue, and certain meditation states. Used in attention research — elevated TAR correlates with impaired response inhibition.; Putman et al. (2010)
Theta/Beta Ratio TBR Metric · Cross-Band Ratios: θ / β; The most-studied EEG biomarker for ADHD. Elevated TBR was FDA-cleared in 2013 as a diagnostic aid for ADHD in children (NEBA system). Higher values = reduced cortical arousal. Used alongside clinical assessment, never as a standalone diagnostic.; Monastra et al. (1999); Angelidis et al. (2016)

U

UMAP UMAP Machine Learning: Uniform Manifold Approximation and Projection — a manifold learning algorithm that reduces high-dimensional data to 2D or 3D while preserving local and global structure. NeuroSkill™ runs UMAP on GPU via a CubeCL backend, projecting thousands of 32-D brain-state embeddings into an interactive 3D point cloud viewable in the UMAP window.

W

Wakefulness Metric · Consciousness Metrics: 0.40·BAR + 0.35·(1−TAR) + 0.25·(1−drowsiness); Wakefulness level — inverse drowsiness modulated by the Beta/Alpha Ratio (BAR) and Theta/Alpha Ratio (TAR). High values indicate an alert, active brain state; low values suggest drowsiness or sleep onset. Grounded in alpha/theta arousal indices.; Klimesch (1999)
WebSocket Protocol: A full-duplex, persistent TCP protocol (RFC 6455) enabling real-time bidirectional communication between NeuroSkill™ and external clients. NeuroSkill™ listens on port 8765 and registers via mDNS as `_skill._tcp`. Clients send JSON command messages; NeuroSkill™ responds with JSON data and pushes event streams (EEG bands, device status, labels) continuously.
Welch Method Signal Processing: A PSD estimation technique that averages the periodograms of multiple overlapping, windowed segments of the signal. Reduces the variance of PSD estimates compared to a single-window FFT, at the cost of frequency resolution. NeuroSkill™ uses 4 overlapping 512-sample Hann-windowed segments per epoch, striking a balance between resolution and variance.
wgpu Framework: A cross-platform GPU compute API in Rust, implementing the WebGPU standard. Used by NeuroSkill™ for the FFT pipeline and ZUNA encoder compute shaders. Targets Metal on macOS, Vulkan on Linux/Windows, and DX12 on Windows. Achieves ~125 ms end-to-end latency for all 60+ metrics on consumer hardware.

Z

ZUNA Encoder ZUNA NeuroSkill™: NeuroSkill™'s on-device neural network for computing brain-state embeddings. Maps each 5-second EEG epoch (4 channels × 1280 samples) to a 32-dimensional float vector via GPU compute shaders. Enables semantic similarity search across sessions without sending raw EEG to any external service. Architecture: lightweight 1D convolutional network.

Metric references link directly to the primary peer-reviewed source via DOI. For metrics without a citable DOI, the link goes to the full reference list. All definitions describe NeuroSkill™'s implementation and are not intended as clinical guidance.

← Back to NeuroSkill™