EchoDepth Insights
Research and analysis on emotion recognition AI, FACS methodology, polygraph science, human reliability assessment, and defence security — from the team building the UK's only purpose-built classified emotion AI platform.
Emotion AI for Defence
Eight foundational articles covering the scientific and operational basis for camera-based emotion recognition in defence and security environments.
What Is FACS? Facial Action Coding System Explained
FACS defines 44 Action Units mapped to specific facial muscles. The only peer-reviewed framework that produces anatomy-grounded, legally defensible emotion output.
Credibility AssessmentPolygraph False Negative Rate: Why the Science Doesn't Support Security Screening
The NAS 2003 review found false negative rates reaching 47%. Courts won't accept it. Procurement bodies are moving on. The evidence, examined.
Emotion AIVAD Model Explained: Valence, Arousal & Dominance in Emotion AI
Labelling a face as "angry" discards most of what the face is actually communicating. VAD represents emotional state in three continuous dimensions.
Defence OperationsHuman Reliability Assessment in Defence: The Case for Real-Time Monitoring
Fatigue is implicated in 21–23% of major military incidents. HRA evaluates it at a point in time. Real-time monitoring closes the gap.
Insider ThreatInsider Threat Detection: Why Emotional Signals Precede the Digital Footprint
Carnegie Mellon CERT found an average 14-month gap between first observable precursor and first malicious act. The signals were there. The monitoring wasn't.
Operator ReadinessOperator Fatigue Monitoring: What It Is, How It Works, and Why It Matters
Fatigue implicated in 21–23% of major military incidents. Camera-based FACS detection without wearables or self-report. How AU46 and VAD arousal track fatigue onset.
DeploymentSCIF-Compatible AI Deployment: Requirements, Standards, and What to Look For
"SCIF-compatible" is a specific technical requirement — zero external transmission, no cloud, all inference local. What to verify in defence AI procurement.
Cyber SecurityDeepfake Detection Technology: How AU Temporal Coherence Identifies Synthetic Video
Deepfakes synthesise pixels. They cannot synthesise neuromuscular timing. POKERFACE detects synthetic video through Action Unit temporal coherence analysis.
ScienceDimensional Emotion Models in Defence: Why VAD Outperforms Discrete Labels
Categorical emotion labels force a choice where the evidence is continuous. Why dimensional models — VAD — are the correct framework for defence and security emotion AI.
Use CaseOperator Fatigue Monitoring: UAS, SOC, and SCIF Deployment
Real-world deployment of FACS-based fatigue detection in UAS operations, security operations centres, and SCIF environments. Technical specs and integration requirements.