Multimodal stress check

Understandyour stress,gently.

SafeSpace reads three signals: your body, your voice and your own words. Then it explains what it found in plain language.

Start the check

BodyVoiceWords

A short breathing pause, whenever you want one.

signals read together: 3; signals read together
short questions: 7; short questions
explanations with every result: SHAP; explanations with every result

(01) What it is

Stress isn’t one number. SafeSpace listens to your body, your voice and your own words, then reads them together for a fuller picture than any one signal can give.

Privacy first

Edge processing keeps your data secure.

Real-time

Stress levels are estimated as soon as your data arrives.

(02) How it works

Three signals,
one reading.

Each signal is scored by its own model. A late-fusion model then weighs them into a single, explained result.

01
Physiological signals
EDA, ECG, temperature and accelerometry, modelled on the WESAD dataset, to pick up autonomic arousal your mind might not notice yet.
CSV upload
02
Self-assessment
Seven stress items from the DASS-21 questionnaire add your own account of how things feel, which no sensor can measure.
7 questions
03
Voice analysis
Emotional cues in speech, such as pitch jitter and harmonic patterns, learned from the RAVDESS and IEMOCAP datasets.
Voice recording

(03) Explainable by design

No black box.Every reading shows its reasons.

SafeSpace doesn’t just apply explainable AI. It compares techniques to find which best fits a system using several kinds of data, so you can see which signals moved your result and by how much.

SHAPSHapley Additive exPlanations, a game-theory method for estimating feature importance.
LIMELocal, interpretable explanations of each individual decision.
Integrated GradientsAttribution for the deep-learning parts of the pipeline.

Research impact