C3 - Product Architecture Overview

DBT

Linehan, 1993

ACT

Hayes, 1999

Somatic Experiencing

Levine, 1997

Polyvagal Theory

Porges, 2011

Affective Computing

Picard, 1997

James-Lange Theory

James, 1884

🔌 Hardware Layer

🤲

Pressure Sensors

4-point FSR array detects squeeze patterns, intensity, and location

🎙️

Voice Prosody

MEMS mic analyzes pitch, tempo, volume (not words)

👆

Capacitive Touch

8-pad array tracks touch location, duration, stroking patterns

📱

9-axis IMU

Detects movement: shaking, rocking, cradling, stillness

💡

RGB LED Ring

24 addressable LEDs display emotional state (Plutchik wheel)

📳

HD Haptic

LRA provides precise tactile feedback for grounding

🧠 Core Processing Engine

1. Multimodal Sensing

Fuses pressure, voice, touch, movement → Emotional state inference using Affective Computing principles. Context-aware: time, location, history.

↓

2. Emotion Classification

Maps to Plutchik's 8 basic emotions + intensity. Confidence scoring. Tracks transitions and patterns. Personalized baseline learning.

↓

3. Adaptive Intervention

Selects optimal therapy-based intervention (DBT/ACT/SE). Considers: current state, past effectiveness, context, user preference. Guides execution via LED + haptic.

💻 Software Layer

🎙️

Voice Journaling

AI-powered ACT prompts, pattern detection, narrative therapy

📊

Pattern Analysis

Emotion heatmaps, trigger analysis, growth timeline

🧘

Intervention Library

20+ evidence-based practices, personalized recommendations

🤝

Co-Regulation

Pair with loved ones, sync breathing, ambient presence

🔒

Privacy Control

Local-first storage, E2E encryption, granular sharing

🎨

Customization

Emotion colors, squeeze vocabulary, personal rituals

🧘‍♀️

Emotional Regulation

Evidence-based techniques for managing anxiety, stress, anger

🌱

Self-Awareness

Understand emotional patterns, triggers, growth over time

💕

Connection

Strengthen relationships through co-regulation and ambient intimacy

🔐

Dignity & Privacy

Discreet support that honors vulnerability in shared spaces

C4 - Multimodal Emotion Sensing FlowSenze 2.0
Multimodal Emotion Sensing Flow

                        🤲
                        Pressure Input
                    
FSR Sensor Array (4 points)

                            P1 (Top):
                            850 / 1024
                        

                            P2 (Right):
                            720 / 1024
                        

                            P3 (Bottom):
                            680 / 1024
                        

                            Pattern:
                            Tight Squeeze
                        
→ Pressure Analysis

                        Classifies squeeze type (hug, tight squeeze, pat, stroke) and calculates intensity. High pressure (>700/1024) + bilateral pattern = stress indicator.
                    

                        avg_pressure = 762.5

                        pattern = "bilateral_tight"

                        stress_score = 0.85
                    
Pressure Signal
High Stress
Confidence: 85%

                        🎙️
                        Voice Prosody
                    
MEMS Microphone

                            Pitch:
                            280 Hz (elevated)
                        

                            Speech Rate:
                            180 wpm (fast)
                        

                            Volume:
                            65 dB
                        

                            Tremor:
                            Detected
                        
→ Prosody Analysis

                        Analyzes pitch contour, speech rate, and voice tremor. No word recognition. Elevated pitch + fast rate + tremor = anxiety marker.
                    

                        pitch_z_score = +1.8

                        rate_percentile = 92nd

                        anxiety_score = 0.78
                    
Voice Signal
Anxious
Confidence: 78%

                        👆
                        Touch Pattern
                    
Capacitive Touch (8-pad)

                            Active Pads:
                            1, 2, 3 (sequence)
                        

                            Duration:
                            0.8s per touch
                        

                            Frequency:
                            3x in 10s
                        

                            Gesture:
                            Rapid stroking
                        
→ Touch Classification

                        Identifies gesture type and calculates touch frequency. Rapid repeated stroking = self-soothing attempt, confirms distress state.
                    

                        gesture = "rapid_stroke"

                        touch_frequency = 0.3 Hz

                        soothing_attempt = true
                    
Touch Signal
Self-Soothing
Confidence: 72%

                        📱
                        Movement Data
                    
9-axis IMU

                            Accel X:
                            ±2.5g
                        

                            Accel Y:
                            ±1.8g
                        

                            Pattern:
                            Oscillating (2Hz)
                        

                            Type:
                            Rocking motion
                        
→ Movement Classification

                        Detects rocking, shaking, or stillness. Rhythmic rocking = somatic self-regulation behavior, common in anxiety.
                    

                        movement_type = "rhythmic_rock"

                        frequency = 2.1 Hz

                        regulation_attempt = true
                    
Movement Signal
Active Regulation
Confidence: 80%

                🧭
                Contextual Factors
            
Time
3:15 PM
Location
Work
History
3 PM stress pattern
Sleep
6.2 hrs last night
Anxiety (Moderate-High)
Intensity:
75%
Confidence:
82%
Recommended Intervention
Paced Breathing (DBT)
85% effective for you in past

C5 - Intervention Decision Tree

🧠 Emotional State Detected

Multimodal sensing → Emotion + Intensity + Confidence

↓

😊 Ventral Vagal

Safe & Social State
Calm, connected, open to engagement

Goal: Maintain & Savor

😰 Sympathetic

Fight or Flight
Anxious, stressed, elevated arousal

Goal: Downregulate

😶 Dorsal Vagal

Freeze / Shutdown
Dissociated, numb, withdrawn

Goal: Gentle Activation

↓ ↓ ↓

Gratitude Moment Pos Psych

Voice prompt: "What's one good thing right now?" Record & save as resource.

⏱️ 2 min

★★★★★ 92% for maintenance

Savoring Exercise Mindfulness

LED warm glow. "Notice 3 sensations in your body right now."

⏱️ 3 min

★★★★☆ 88% for calm states

Connection Pulse Polyvagal

If paired: Send "thinking of you" pulse. LED gentle pink wave.

⏱️ 1 min

★★★★★ 95% when connected

Paced Breathing DBT-TIPP

4-7-8 pattern. Inhale 4s, Hold 7s, Exhale 8s. Haptic + LED guide rhythm.

⏱️ 5 min

★★★★★ 85% for you historically

5-4-3-2-1 Grounding DBT

LED prompts: "Name 5 things you see..." Voice captures responses. Anchors to present.

⏱️ 3 min

★★★★☆ 78% effective

Progressive Muscle Relax SE

Voice guides: "Squeeze plush tight... now release." LED tenses (bright) → relaxes (dim).

⏱️ 8 min

★★★★☆ 72% effective

Leaves on Stream ACT

Visualize thoughts as leaves floating away. LED shows "thought" drifting off. Cognitive defusion.

⏱️ 4 min

★★★☆☆ 65% (new for you)

Shake It Off SE-TRE

Prompt physical release: shake plush vigorously. LED responds to movement. Discharge trapped energy.

⏱️ 2 min

★★★★☆ 80% for acute stress

"I'm Here" Pulse Polyvagal

Gentle haptic reminder. LED soft breathing. "No need to do anything. Just noticing you."

⏱️ Ongoing

★★★★★ 90% as first step

Micro-Movement SE

Tiny invitation: "Can you squeeze me once?" Celebrate any response. Gradual re-engagement.

⏱️ 1 min

★★★★☆ 75% gets response

Reach Out Prompt Safety Net

"Would you like to text someone?" Pre-written options to emergency contact. OR decline, presence is enough.

⏱️ 2 min

★★★★☆ 70% if offered gently

🎯 Selection Algorithm Considers:

📊

Past Effectiveness

🕐

Time of Day

📍

Location Context

🔄

Recent Usage (Vary)

❤️

User Preference

Hardware Architecture - Technical Exploded View

Fabric Exterior Shell

Material: Premium cotton blend, washable

Soft woven textile providing tactile comfort. Beige colorway matches plush aesthetic. Integrated with capacitive touch grid.

🧵

RGB LED Ring Array

IC: WS2812B, 24 LEDs, 16.7M colors

Addressable LEDs in circular formation. Individual control for smooth animations. Displays Plutchik emotion wheel colors. PWM brightness 0-100%.

💡

Main Circuit Board (PCB)

MCU: nRF52840 (ARM Cortex-M4 64MHz)

Central processor coordinating all functions. BLE 5.2, 1MB flash, 256KB RAM. Secure element (ATECC608) for encryption keys. On-device TinyML inference.

⚡

Rechargeable Battery

Type: LiPo 800mAh @ 3.7V (2960mWh)

12-day battery life typical use. USB-C fast charge (30min = 50%, 1hr = 100%). Protection circuit: over-charge, over-discharge, short-circuit.

🔋

HD Haptic Engine

Type: Linear Resonant Actuator (LRA)

Precise tactile feedback 50-300Hz. Response time <5ms. Simulates heartbeat, breathing patterns, alerts. Power efficient <100mW active.

📳

Sensor Array Module

4× FSR + 8× Touch + IMU + MEMS Mic

Pressure sensors (0-10N), capacitive touch pads, 9-axis IMU (accel+gyro+mag), directional microphone (cardioid, 65dB SNR). Physical mic disable switch.

📡

Modular Clip Mechanism

Material: Recycled ABS, spring-loaded

Interchangeable attachment system. Options: spring clip, velcro strap, safety pin, magnetic mount. Universal compatibility with plush toys.

📎

Ø 2.8" × 0.6"

Diameter × Thickness

Weight: 52g (light enough for any plush)

🧠 Processing & Connectivity

Main MCU

nRF52840 (ARM Cortex-M4)

Clock Speed

64 MHz

Flash Memory

1 MB (firmware + ML models)

RAM

256 KB (processing buffer)

Local Storage

8 GB eMMC (journal, patterns)

Bluetooth

BLE 5.2 (30ft range, low energy)

Encryption

ATECC608 secure element (AES-256)

📊 Sensor Specifications

Pressure (FSR)

4 points, 0-10N, 10-bit (1024)

Microphone

MEMS, 100Hz-10kHz, SNR 65dB

Touch

8 capacitive pads, <5mm resolution

IMU

9-axis (accel ±8g, gyro ±2000°/s)

Sampling Rate

100 Hz continuous

⚡ Power Budget (12-day battery life)

Mode	Power	Daily
Idle (22h)	5 mW	110 mWh
Active Sensing (1.5h)	50 mW	75 mWh
LED Display (30min)	100 mW	50 mWh
Total Daily		~235 mWh

Battery: 2960 mWh → 12.6 days life ✓

🏭 Manufacturing & Cost

BOM Cost (10K units)

$48 per unit

Target Retail

$99 - $129 (3x margin)

Materials

Recycled ABS, cotton fabric

Certifications

FCC, CE, RoHS compliant

Warranty

1 year + optional 2yr extend