Glossary of
key terms.

FACS — Facial Action Coding System

The scientific standard for objectively measuring facial movement. Developed by Paul Ekman and Wallace Friesen in 1978, FACS decomposes every visible facial expression into discrete, numbered Action Units (AUs). EchoDepth uses FACS as its core analysis framework.

Reference: Ekman, P. & Friesen, W.V. (1978). Facial Action Coding System. Consulting Psychologists Press.

Action Unit (AU)

A numbered, discrete facial muscle movement defined by FACS. Each AU represents a specific facial motion — for example, AU1 (inner brow raise), AU4 (brow lowerer), AU12 (lip corner pull). EchoDepth tracks 44 Action Units per frame in real time, scoring each on a continuous 0.0–1.0 intensity scale. Many AUs are involuntary and cannot be consciously suppressed.

See how EchoDepth uses Action Units →

VAD — Valence, Arousal, Dominance

A three-dimensional model of emotional state. EchoDepth maps AU activations to continuous VAD scores (0.0–1.0 per dimension):

• Valence — positive (1.0) to negative (0.0) emotional state. Low valence indicates distress or anxiety.
• Arousal — calm (0.0) to high-activation (1.0). Elevated arousal may indicate stress, fear, or urgency.
• Dominance — submissive/vulnerable (0.0) to in-control (1.0). Low dominance indicates powerlessness.

See full VAD methodology →

FCA Consumer Duty

A UK regulation (effective July 2023) that requires FCA-regulated financial services firms to: (1) identify vulnerable customers, (2) take appropriate action to avoid foreseeable harm, and (3) document that vulnerability was considered before consequential decisions. EchoDepth directly addresses all three requirements through real-time vulnerability detection and timestamped audit trails.

Read: Why vulnerability detection needs to be real-time →

Vulnerability Detection

The identification of emotional distress, cognitive load, or situational vulnerability in customer interactions — typically in real time. In financial services, vulnerability detection is a regulatory requirement under FCA Consumer Duty. EchoDepth detects vulnerability through voice prosody analysis (phone calls), facial AU analysis (video), and linguistic markers (text), classifying vulnerability into tiers from standard to crisis.

See vulnerability tier classification →

Cognitive Load

The amount of mental processing capacity being used during a task or interaction. In EchoDepth, elevated cognitive load is detected through sustained high arousal, specific AU patterns (AU1+4 brow tension), increased micro-pauses in speech, and speech rate deceleration. High cognitive load during financial interviews may indicate processing difficulty, constructed responses, or emotional overload.

Temporal Coherence

The consistency of emotional signals across a session, as opposed to single-frame analysis. EchoDepth scores temporal coherence by establishing a baseline from the opening minutes of each session and tracking how VAD scores and AU activations evolve over time. Sudden divergence from established patterns is flagged for review.

Prosody (Voice Analysis)

The patterns of pitch, rhythm, rate, energy and timing in speech — independent of the words being spoken. EchoDepth's voice module analyses prosodic features to detect vocal stress markers, micro-pauses, pitch collapse, and vocal tremor. This is the primary modality for phone-based vulnerability detection in debt collections and complaint handling.

Read: Voice vulnerability detection in collections →

Sentiment Analysis (vs. Emotional AI)

Traditional sentiment analysis classifies words and phrases as positive, negative or neutral. It reads the script. Emotional AI (as implemented by EchoDepth) reads the performance — analysing involuntary physiological signals (facial AUs, voice prosody) that are independent of what the person says and largely immune to conscious manipulation. This distinction is critical in financial services, where polite language may mask genuine distress.

Read: Why sentiment analysis fails for financial services →

GDPR — General Data Protection Regulation

The EU and UK regulation governing data protection and privacy. EchoDepth's privacy architecture is structurally GDPR-compliant: no raw video is retained, no biometric data is stored, and only anonymised VAD scores and AU activations are output. Edge deployment is available for organisations where data cannot leave the premises.

E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness

Google's framework for evaluating content quality. EchoDepth content is authored by Jonathan Prescott (CEO, Cavefish) — MBA from Bayes Business School, former Director of Digital at The Royal Mint, Director of Digital Performance at Assurant, and NYU Stern School of Business instructor. Cavefish is an FCA Regulatory Sandbox participant with 13 industry awards.