Cognitum/Emotion Detection 0.6.0 EXPERIMENTAL

Emotion Detection

Infers emotional states — arousal, stress, calm — from combined gait analysis and breathing pattern features. Research build, ESP32-S3.
Current emotion state
CALM
stable · last update —
Confidence
0.78

Live signal metrics

Arousal level
—
0.0 – 1.0 range
Stress score
—
windowed avg
Calm index
—
parasympathetic
Valence
—
negative ↔ positive
Breathing rate
—
breaths / minute
Input pipeline
CSI motion + breathing fusion
streaming
Hardware
ESP32-S3
Sample rate
50 Hz
Window
8.0 s
Stride
1.0 s
Recent emissions
last 60 seconds
emotion_state
—
arousal_level
—
stress_score
—
valence
—

Recent events

TimeEventValueConfidence

Live inference

Real-time arousal, stress and calm gauges. Streaming gait + breathing features at 50 Hz, inference window 8 s.
Arousal
0.00
level
σ 0.00 · 30s
Stress
0.00
score
σ 0.00 · 30s
Calm
0.00
index
σ 0.00 · 30s
Valence
0.00
−1 ↔ +1
σ 0.00 · 30s
CSI motion · gait
50 Hz · channel 0
rms 0.000stride 1.12 s
Breathing pattern
50 Hz · band 0.1–0.7 Hz
— bpmi:e 1 : 1.5

Inference trace

last 30 windows · 1 s stride
arousal stress calm

Events

Live emission log — emotion_state, arousal_level, stress_score, valence. Filter by type or export.
Time Event Value Confidence Source

Input signals

CSI motion features for gait analysis and band-passed respiration envelope. Both streams fused into the inference window.
CSI motion (gait)
Wi-Fi channel state information, 56 subcarriers
locked
Sample rate
50 Hz
Subcarriers
56
Stride period
1.12 s
Cadence
107 spm
Gait variance
0.043
RSSI
−42 dBm
Breathing
Band-passed 0.1–0.7 Hz from CSI doppler envelope
locked
Rate
15.2 bpm
Depth
0.62
I:E ratio
1 : 1.5
HRV proxy
41 ms
SNR
18.4 dB
Window
8.0 s

Live waveforms

Gait envelope
50 Hz · 8.0 s
Breathing envelope
50 Hz · 8.0 s

Extracted features

FeatureChannelValueUpdated

Model

Inference model metadata, weights, and runtime stats.
Experimental research build. Inferences are illustrative and not clinically validated. Do not use for medical decisions.
Identity
cogs/emotion-detect
research experimental hard
Version
0.6.0
Size
30 KB
Hardware
ESP32-S3
Difficulty
hard
Source
cogs/src/cogs/emotion-detect
Status
loaded
Emitted events
topics the cog publishes
EventTypeRate
emotion_stateenum1 / window
arousal_levelfloat1 Hz
stress_scorefloat1 Hz
valencefloat1 Hz

Runtime

Inference latency
—
ms / window
Windows / min
—
stride 1 s
Mean confidence
—
rolling 5 min
Memory
30 KB
on-device

Class distribution (last 5 min)

Calibration

Capture a personal baseline so arousal, stress and calm scores are relative to your resting state.
Current baseline
Last captured —
active
Resting BPM
14.3
Resting HRV
52 ms
Cadence
98 spm
Gait var
0.038
Thresholds
override class boundaries

Settings

Hardware bindings, runtime parameters, and emission targets for the emotion detection cog.
Hardware
target device binding
Runtime
windowing + inference
Emission
events published by the cog
enum: calm · neutral · aroused · stressed
continuous float 0.0 – 1.0
continuous float 0.0 – 1.0
continuous float −1.0 – +1.0
Privacy
on-device data handling
no raw signals leave the ESP32-S3
ring buffer, 24 h retention
opt-in research telemetry
Saved