← Back to CAIT Orionus Logo Orionus

Behavioral Intelligence for Proactive Security

Security-focused multimodal emotion and intent recognition for public spaces. 4 modalities, 10 security intents, 5 threat levels, deception detection, and zone-based monitoring.

Try Live Demo →

Live Demo

Try Orionus directly in your browser. Register with your email for a 60-second trial session.

Having issues? Open demo in a new tab →

4 Modalities
10 Security Intents
5 Threat Levels
16+ Fuzzy Rules

Four Modalities

Comprehensive behavioral analysis through face, voice, text, and full-body movement signals.

🧑

Facial Analysis

Emotion detection, micro-expression analysis, and gaze tracking. Identifies stress indicators, concealed emotions, and deceptive expressions in real-time.

🎙

Voice Analysis

Speech emotion recognition with voice stress analysis. Detects aggression, nervousness, deception cues, and emotional arousal from acoustic features.

📝

Text Analysis

Real-time linguistic analysis of spoken content. Identifies threatening language, evasive patterns, inconsistencies, and emotional state from word choice.

🏃

Movement Analysis

Full-body pose estimation via MediaPipe 33 landmarks. Tracks gait, posture, gestures, pacing, and spatial behavior for behavioral profiling.

10 Security Intents

Orionus classifies observed behavior into 10 security-relevant intent categories.

Intent Description Threat Association
NormalStandard, expected behavior with no indicators of concernNone
AgitatedVisible signs of distress, anxiety, or emotional arousalLow
AggressiveHostile body language, raised voice, confrontational stanceElevated
Attack IntentClear indicators of imminent physical aggression or violenceCritical
ConcealmentDeliberate hiding of objects, face, or identityElevated
EvasionActive avoidance of cameras, personnel, or checkpointsElevated
LoiteringProlonged presence without clear purpose in sensitive areasLow
PanicFlight response, extreme fear, or crowd panic indicatorsHigh
DeceptiveMulti-channel deception signals across face, voice, and textHigh
ErraticUnpredictable, disoriented, or incoherent behaviorElevated

5 Threat Levels

Graduated threat assessment for proportional security response.

NONE

Normal behavior, no intervention needed

LOW

Minor indicators, passive monitoring

ELEVATED

Active monitoring, personnel alert

HIGH

Immediate personnel dispatch

CRITICAL

Emergency response activation

Fuzzy Rule Base

16+ rules across 5 categories govern the security inference engine.

🛡

Aggression Rules

Rules mapping facial anger, vocal aggression, hostile language, and aggressive posture to threat levels and security intents.

🔍

Concealment Rules

Rules detecting deliberate hiding behaviors, identity concealment, and evasive movement patterns in security zones.

🎭

Deception Rules

Cross-modal conflict rules identifying mismatches between face, voice, and text signals that indicate deceptive behavior.

Panic Rules

Rules for detecting mass panic, flight responses, and extreme fear indicators across all four modalities.

📍

Zone-Sensitive Rules

Rules that adjust sensitivity based on zone type and context — higher sensitivity in restricted areas, lower in public concourses.

Deception Detection

5-channel deception analysis for comprehensive veracity assessment.

🧑

Micro-Expressions

Fleeting facial expressions that reveal concealed emotions

👀

Gaze Patterns

Eye movement anomalies associated with deception

🎙

Voice Stress

Vocal micro-tremors and pitch shifts under cognitive load

💬

Linguistic Cues

Hedging, distancing, and inconsistency in verbal content

🧍

Body Cues

Self-soothing, fidgeting, and postural shifts

Zone Monitoring

10 zone types with context-specific sensitivity multipliers.

Zone Type Example Sensitivity
Security CheckpointAirport gates, building entryVery High
Restricted AreaStaff-only, server roomsVery High
Boarding AreaGate lounges, platformsHigh
Baggage ClaimCarousel areas, pickup zonesElevated
Transit CorridorHallways, tunnels, walkwaysElevated
Public ConcourseMain halls, food courtsStandard
Retail ZoneShops, duty-free areasStandard
Parking AreaGarages, drop-off zonesElevated
VIP LoungePremium waiting areasHigh
Exterior PerimeterBuilding perimeter, fencesElevated

Movement Analysis

MediaPipe 33-landmark pose estimation powers 10 behavioral movement features.

10 Movement Features

Gait speed, direction changes, arm swing amplitude, postural sway, personal space violations, pacing frequency, hand-to-face contact, sudden movements, crouching/ducking, and spatial clustering.

Highest Weight: 35%

Movement modality carries the highest weight in security context (35%), reflecting the importance of behavioral body language in threat detection. Face (25%), voice (20%), and text (20%) complete the fusion.

Alert Engine

Intelligent alert management with 4 priority levels, cooldown, and escalation logic.

🔔

4 Priority Levels

Informational, Warning, Urgent, and Critical alert levels. Each triggers different notification channels and response protocols.

Smart Cooldown

Prevents alert fatigue with intelligent cooldown periods. Same-subject alerts are suppressed within configurable time windows.

📈

Escalation Logic

Automatic escalation when threat indicators persist or intensify. Low-level alerts upgrade based on duration and pattern matching.

📋

Audit Trail

Full logging of all alerts, responses, and outcomes. Complete audit trail for compliance, review, and system optimization.

Use Cases

Airports

Security checkpoints, boarding areas, baggage claim, and terminal monitoring with zone-sensitive threat assessment.

🚉

Transit Stations

Metro, rail, and bus terminal security with crowd monitoring, platform surveillance, and corridor analysis.

🏟

Venues

Stadiums, concert halls, and large events with crowd behavior analysis, entry screening, and perimeter monitoring.