Twenty-nine sensors work together to distinguish between safe play and real danger — swimming vs drowning, roughhousing vs assault.
"Imagine 29 tiny helpers on your wrist — one listens, one feels your heartbeat, one checks the temperature, one knows if you're in water. They all talk to each other to figure out if you're having fun or if something is wrong."
SafeStep integrates 29 distinct sensor inputs: 6-axis IMU (accelerometer + gyroscope), PPG heart rate, BIA hydration electrodes, NIR optical sensor, GSR stress electrodes, MEMS microphone array, capacitive fingerprint scanner, moisture detection electrodes, temperature sensor, barometric pressure, UV index, ambient light, and multiple connectivity radios.
The sensor fusion engine runs a hierarchical decision tree: first classifying the general context (indoor/outdoor, moving/stationary, wet/dry), then narrowing to specific scenarios using weighted sensor combinations.
Key differentiations: Swimming vs Drowning (motion pattern + heart rate + submersion depth), Bed Wetting vs Showering (time of day + location + water type), Hot Car vs Outdoor Play (temperature curve + barometric pressure + motion), Roughhousing vs Assault (audio + heart rate + GSR + motion intensity).
Each sensor has a reliability weight that adjusts dynamically. If one sensor is occluded or noisy, the system redistributes confidence to remaining sensors rather than producing false alerts.
| Total Sensors | 29 distinct inputs |
| Accelerometer | ±16g, 3-axis |
| Gyroscope | ±2000°/s, 3-axis |
| Heart Rate | PPG, ±2 BPM, continuous |
| Hydration | BIA + NIR, ±3% accuracy |
| Audio | Dual MEMS microphone array |
| Temperature | ±0.1°C, skin + ambient |
| Moisture | 8-point conductivity grid |
| Barometric | ±0.1 hPa, altitude detection |
| Fusion Rate | 100Hz sensor polling, 10Hz classification |
No single sensor can tell the difference between drowning and swimming. But when you combine heart rate, blood oxygen, motion patterns, water pressure, skin conductance, and ambient sound — the picture becomes unmistakable. This is sensor fusion.
Raw sensor data flows through 6 stages in under 12 milliseconds — from electrical signals to life-saving decisions.
28 sensors sampling at 1,240 samples/sec combined
Noise filtering, normalization, feature extraction
Time of day, location history, activity baseline, guardian proximity
Multi-layer neural network comparing 200+ signal combinations
Scenario probability calculated with weighted sensor inputs
Alert level determined: Monitor → Notify → Warn → SOS
Click any scenario to see exactly which sensors fire, what they read, and how the ML model decides between danger and safety — in real time.
How SafeStep knows the difference in under 8 seconds
Child is submerged and in distress — immediate SOS cascade triggered
Child is actively swimming in a supervised environment — monitoring continues normally
| Sensor | Drowning | Swimming | Critical? |
|---|---|---|---|
SpO2 | Drops below 92% rapidly | Stays 94-98% (breath-holding dips briefly) | YES |
Heart Rate | Spikes to 150+ BPM then drops suddenly | Elevated 100-130 BPM (exercise zone) | YES |
Accelerometer | Erratic thrashing → sudden stillness | Rhythmic arm stroke patterns detected | YES |
Gyroscope | Random rotations, no pattern | Consistent rotation matching swim strokes | Supporting |
Skin Moisture | 100% (full submersion) | 100% (full submersion) — same reading | Supporting |
Barometric Pressure | Increases (depth > 30cm) | Surface-level pressure or slight dips | YES |
Ambient Light | Drops dramatically (underwater) | Fluctuates (surface splashing) | Supporting |
GSR | Massive spike (panic response) | Moderate elevation (exertion) | YES |
Microphone | Gurgling/silence after cry | Splashing sounds, laughter | Supporting |
GPS | Stationary in water body | Moving within pool/beach zone | Supporting |
When SpO2 drops below 92% AND heart rate shows spike-then-drop pattern AND accelerometer shows erratic-to-still transition AND GSR shows panic spike — drowning confidence > 95%. SOS triggered within 8 seconds.
Complete reference with sampling rates and safety roles
Optical PPG sensor detecting blood volume changes
Calculated from R-R intervals; measures autonomic nervous system balance
Measures skin conductance via electrodes; indicates emotional arousal
Infrared thermopile sensor measuring skin surface temperature
Dual-wavelength LED measuring hemoglobin oxygen saturation
PTT-based estimation using PPG waveform analysis
Capacitive humidity sensor detecting perspiration and skin hydration
Capacitive sensor with 508 DPI, FAR <0.001%
4-electrode bioimpedance at 10 frequencies (DC-100kHz)
A skin moisture reading of 100% means nothing on its own. But combine it with dark ambient light, sleep-level heart rate, room-temperature air, and a lying-down accelerometer pattern — and you have a bed-wetting event with 90% confidence. The same 100% moisture reading with bright light, elevated heart rate, warm ambient temperature, and standing motion? That's just a shower.
SafeStep's ML model evaluates 200+ signal combinations every 12 milliseconds, building a real-time picture of your child's world. It doesn't just read sensors — it understands situations.
AI Danger Prediction
Microphone, pulse, and emotion sensors fuse together in real-time to classify threats before they escalate.
Water & Drowning Detection
Eight-point conductivity sensors detect submersion instantly, distinguishing between splashing, swimming, and drowning.
Pre-Impact Fall Detection
Predictive algorithms detect free-fall signatures before impact, classifying severity from minor tumbles to critical falls.
Ready to protect your child?