Technical Architecture

Five-layer system.
No prior art has combined these.

A layer-by-layer overview of the EMPATHIQ system architecture. This page describes what each layer does and why it matters. The complete algorithm pseudocode, DEH data schema, RDRS computation parameters, and patent claim language are available in the full Technical Briefing Package — provided to qualified partners under NDA.

Patent ApplicationAU2026903561
Independent Claims3 claims
Dependent Claims17 claims
Novel Core ClaimsDEH · RDRS · Pattern Mirror
Processing100% On-Device
Privacy ArchitectureFederated + Differential Privacy
✓ Openly Published System overview · Layer functions · Privacy architecture · Hardware embodiments
🔒 Under NDA Only Algorithm pseudocode · DEH schema · RDRS parameters · All 20 patent claims
01
Layer 1
Multimodal Emotional State Sensing
Wearable Biosignals Voice Acoustics Keyboard Patterns Contextual Signals Transformer Fusion ESV Output

Outputs the Emotional State Vector (ESV) — a continuous three-dimensional representation of the user's current emotional state in Valence, Arousal, and Dominance space (PAD model, Mehrabian & Russell 1974).

Signal Sources

HRV
RMSSD · SDNN · HF power
Primary arousal and autonomic regulation marker. Neurovisceral Integration Model (Thayer & Lane 2009) establishes HRV as biomarker of prefrontal inhibitory control.
EDA / GSR
SCR amplitude · rise time · AUC
Electrodermal activity — sympathetic nervous system activation. Highly sensitive indicator of emotional arousal and stress onset.
BVP
Blood Volume Pulse · PPG
Blood volume pulse and photoplethysmography providing heart rate dynamics and peripheral vascular tone indicators.
Voice Acoustics
F0 · Jitter · Shimmer · MFCC
Fundamental frequency, speaking rate, amplitude variability processed on-device. No raw audio stored or transmitted.
Keyboard / Touch
WPM · IKI · Error rate
Typing speed, inter-keystroke interval, error rate, navigation depth. Captures arousal-driven behavioural changes without accessing message content.
Context
Time · App · Location
Time of day, day of week, application category, calendar context (with consent), geographic location token (with consent).

ESV Fusion Architecture

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →
TYPE html> Technology — EMPATHIQ.tech
Technical Architecture

Five-layer system.
No prior art has combined these.

A layer-by-layer overview of the EMPATHIQ system architecture. This page describes what each layer does and why it matters. The complete algorithm pseudocode, DEH data schema, RDRS computation parameters, and patent claim language are available in the full Technical Briefing Package — provided to qualified partners under NDA.

Patent ApplicationAU2026903561
Independent Claims3 claims
Dependent Claims17 claims
Novel Core ClaimsDEH · RDRS · Pattern Mirror
Processing100% On-Device
Privacy ArchitectureFederated + Differential Privacy
✓ Openly Published System overview · Layer functions · Privacy architecture · Hardware embodiments
🔒 Under NDA Only Algorithm pseudocode · DEH schema · RDRS parameters · All 20 patent claims
01
Layer 1
Multimodal Emotional State Sensing
Wearable Biosignals Voice Acoustics Keyboard Patterns Contextual Signals Transformer Fusion ESV Output

Outputs the Emotional State Vector (ESV) — a continuous three-dimensional representation of the user's current emotional state in Valence, Arousal, and Dominance space (PAD model, Mehrabian & Russell 1974).

Signal Sources

HRV
RMSSD · SDNN · HF power
Primary arousal and autonomic regulation marker. Neurovisceral Integration Model (Thayer & Lane 2009) establishes HRV as biomarker of prefrontal inhibitory control.
EDA / GSR
SCR amplitude · rise time · AUC
Electrodermal activity — sympathetic nervous system activation. Highly sensitive indicator of emotional arousal and stress onset.
BVP
Blood Volume Pulse · PPG
Blood volume pulse and photoplethysmography providing heart rate dynamics and peripheral vascular tone indicators.
Voice Acoustics
F0 · Jitter · Shimmer · MFCC
Fundamental frequency, speaking rate, amplitude variability processed on-device. No raw audio stored or transmitted.
Keyboard / Touch
WPM · IKI · Error rate
Typing speed, inter-keystroke interval, error rate, navigation depth. Captures arousal-driven behavioural changes without accessing message content.
Context
Time · App · Location
Time of day, day of week, application category, calendar context (with consent), geographic location token (with consent).

ESV Fusion Architecture

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

Missing Modality Handling

The system degrades gracefully — no single modality is required. Masked self-attention in the Transformer handles absent modalities. A confidence field in the ESV output reflects the number and quality of active modalities, enabling downstream components to weight estimates appropriately.

02
Layer 2 · Novel Claim
Personal Emotional Baseline & Decision-Emotion History
★ DEH — Core Novel Claim PEB — Individual Baseline EMA Methodology Online Learning

The DEH is the central novel contribution of this invention. No prior art patent creates or maintains a longitudinal record linking emotional states to decision events and their outcomes.

Personal Emotional Baseline (PEB)

The PEB is constructed exclusively from each user's own longitudinal ESV data — not population norms. This is a key differentiator from all prior art. An anxious person's "calm" looks different from a calm person's "calm." Prior art systems apply population thresholds to everyone.

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

Decision-Emotion History (DEH) — Core Novel Claim

The DEH is the longitudinal personalised dataset maintained by the system, linking each decision event to the emotional state preceding it and the outcome that followed. It is the training foundation for the RDRS model.

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

🔒 All DEH records stored on-device only · Encrypted at rest · User-deletable with immediate effect

03
Layer 3 · Novel Claim
Rash Decision Risk Score Engine
★ RDRS — Core Novel Claim Personalised LSTM Transformer Prior Contextual Risk Factors

The RDRS is a scalar [0,1] representing the probability this specific individual is about to make a decision they will regret, based exclusively on their own historical patterns — not a population model.

RDRS Computation Algorithm

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →
🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →
TYPE html> Technology — EMPATHIQ.tech
Technical Architecture

Five-layer system.
No prior art has combined these.

A layer-by-layer overview of the EMPATHIQ system architecture. This page describes what each layer does and why it matters. The complete algorithm pseudocode, DEH data schema, RDRS computation parameters, and patent claim language are available in the full Technical Briefing Package — provided to qualified partners under NDA.

Patent ApplicationAU2026903561
Independent Claims3 claims
Dependent Claims17 claims
Novel Core ClaimsDEH · RDRS · Pattern Mirror
Processing100% On-Device
Privacy ArchitectureFederated + Differential Privacy
✓ Openly Published System overview · Layer functions · Privacy architecture · Hardware embodiments
🔒 Under NDA Only Algorithm pseudocode · DEH schema · RDRS parameters · All 20 patent claims
01
Layer 1
Multimodal Emotional State Sensing
Wearable Biosignals Voice Acoustics Keyboard Patterns Contextual Signals Transformer Fusion ESV Output

Outputs the Emotional State Vector (ESV) — a continuous three-dimensional representation of the user's current emotional state in Valence, Arousal, and Dominance space (PAD model, Mehrabian & Russell 1974).

Signal Sources

HRV
RMSSD · SDNN · HF power
Primary arousal and autonomic regulation marker. Neurovisceral Integration Model (Thayer & Lane 2009) establishes HRV as biomarker of prefrontal inhibitory control.
EDA / GSR
SCR amplitude · rise time · AUC
Electrodermal activity — sympathetic nervous system activation. Highly sensitive indicator of emotional arousal and stress onset.
BVP
Blood Volume Pulse · PPG
Blood volume pulse and photoplethysmography providing heart rate dynamics and peripheral vascular tone indicators.
Voice Acoustics
F0 · Jitter · Shimmer · MFCC
Fundamental frequency, speaking rate, amplitude variability processed on-device. No raw audio stored or transmitted.
Keyboard / Touch
WPM · IKI · Error rate
Typing speed, inter-keystroke interval, error rate, navigation depth. Captures arousal-driven behavioural changes without accessing message content.
Context
Time · App · Location
Time of day, day of week, application category, calendar context (with consent), geographic location token (with consent).

ESV Fusion Architecture

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

Missing Modality Handling

The system degrades gracefully — no single modality is required. Masked self-attention in the Transformer handles absent modalities. A confidence field in the ESV output reflects the number and quality of active modalities, enabling downstream components to weight estimates appropriately.

02
Layer 2 · Novel Claim
Personal Emotional Baseline & Decision-Emotion History
★ DEH — Core Novel Claim PEB — Individual Baseline EMA Methodology Online Learning

The DEH is the central novel contribution of this invention. No prior art patent creates or maintains a longitudinal record linking emotional states to decision events and their outcomes.

Personal Emotional Baseline (PEB)

The PEB is constructed exclusively from each user's own longitudinal ESV data — not population norms. This is a key differentiator from all prior art. An anxious person's "calm" looks different from a calm person's "calm." Prior art systems apply population thresholds to everyone.

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

Decision-Emotion History (DEH) — Core Novel Claim

The DEH is the longitudinal personalised dataset maintained by the system, linking each decision event to the emotional state preceding it and the outcome that followed. It is the training foundation for the RDRS model.

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →

🔒 All DEH records stored on-device only · Encrypted at rest · User-deletable with immediate effect

03
Layer 3 · Novel Claim
Rash Decision Risk Score Engine
★ RDRS — Core Novel Claim Personalised LSTM Transformer Prior Contextual Risk Factors

The RDRS is a scalar [0,1] representing the probability this specific individual is about to make a decision they will regret, based exclusively on their own historical patterns — not a population model.

RDRS Computation Algorithm

🔒
Full Detail Available Under NDA

The complete algorithm specification, data schema, implementation parameters, and patent claim language for this component are included in the EMPATHIQ Technical Briefing Package — available to qualified partners under a signed non-disclosure agreement.

Request Technical Briefing →
Algorithm 4 — Federated Learning with Differential Privacy
Step 1: Compute local gradient ∇L on device DEH only
Step 2: Clip — ∇L_clip = ∇L · min(1, C / ‖∇L‖₂)
Step 3: Noise — ∇L_dp = ∇L_clip + N(0, σ²·C²) // (ε,δ)-DP satisfied
Step 4: Transmit ∇L_dp only // if user opted into federated learning
Server: θ_new = θ_old − η · mean(∇L_dp across N users)
Hardware Embodiments

Two preferred embodiments — both covered in the patent.

First Preferred Embodiment
Smartphone + Wearable

Smartphone (iOS 16+ / Android 12+, 6GB RAM, NPU/Neural Engine) paired with consumer wrist-worn wearable providing minimum HRV, BVP, and EDA. Compatible with Samsung Galaxy Watch 4+, Fitbit Sense 2+, Empatica EmbracePlus, and equivalents. BLE connection. AES-256 on-device storage. Estimated 5–12% additional battery drain per day.

Second Preferred Embodiment
Smartphone Only

No wearable required. HRV approximated via camera-based rPPG. Arousal proxied via keyboard IKI variance and error rate. Voice, accelerometer, gyro, and context signals identical to first embodiment. ESV has acknowledged lower physiological precision in Arousal dimension. DEH, PEB, RDRS, and intervention layers operate identically. Significantly broader accessible user population.