Your personal emergency response system — from one-tap mobile SOS to standalone IoT crash telemetry with AI-powered impact detection.
// V1 — MOBILE SOS APP
LivePress the EMERGENCY SOS button. Your real-time Google Maps location is sent instantly to your saved contacts via WhatsApp or SMS.
Shake your phone twice. No unlocking needed. The SOS triggers instantly — even when your hands are full or in danger.
Your real-time location is displayed on an always-on dashboard map, so you always know where you are — and your contacts do too.
A built-in national emergency services directory. One tap to dial police, fire, medical — no searching, no delays.
// ADVANCED FEATURES
Turn on extra safety layers from the Advanced Features dashboard.
When your phone hits critical battery, AnomX automatically fires a final SOS with your exact location — before your phone dies.
Once SOS is triggered, AnomX will continuously broadcast your updated GPS coordinates on a timed interval.
Launching Soon// HARDWARE LAYER
◆ Coming SoonAnomX doesn't just run on your phone. An ESP32-based edge node handles real-time impact detection, FFT signal analysis, and BLE communication — all without touching the internet.
Dual-core Xtensa LX6 at 240MHz with integrated WiFi and BLE for real-time edge processing.
Custom GATT services for bidirectional data exchange between ESP32 and the AnomX Android gateway.
Fast Fourier Transform on ADXL345 accelerometer data to detect impact signatures in under 50ms.
A compact ESP32-based wearable node with ADXL345 accelerometer, BLE GATT connectivity, and on-device FFT — pairs directly with the AnomX Android app for real-time impact detection and emergency alerts.
// HOW IT WORKS
Enter the phone number of a trusted contact — family, friend, or colleague.
Select SMS Mode for offline-first sending or WhatsApp for rich media alerts.
Toggle Shake to Alert ON. AnomX is now listening — even when locked.
Press the button or shake twice. Help is notified in under 3 seconds.
// SYSTEM REQUIREMENTS
AnomX V2 places all sensing and decision logic on independent, vehicle-mounted hardware — streaming live telemetry directly to a cloud backend with FFT vibration analysis and Machine Learning crash classification.
// PROJECT OVERVIEW
A standalone IoT-based crash telemetry and cloud dispatch system — a decentralized 'black-box' built on ESP32 that fuses IMU impact data with GPS location, streams live to cloud, and applies FFT + ML to autonomously detect crashes.
Phone-based crash detection depends on OS permissions, phone placement, and battery state. AnomX V2 uses dedicated always-on hardware that works even if the rider's phone is damaged, uncharged, or thrown clear.
Targeted at two-wheeler riders, fleet operators, and emergency-response stakeholders — family contacts, ambulance/dispatch services consuming crash alerts.
Edge node (ESP32 + MPU6050 + GPS) mounts on vehicle. Telemetry goes over Wi-Fi to Oracle Free Tier Cloud, visualized on a remote dashboard accessible anywhere.
V2 — the hardware-independent iteration, moving from phone-only V1 toward a dedicated embedded black-box with edge intelligence.
MPU6050 streams 6-DoF data over I2C; NEO-6M streams NMEA GPS over UART. ESP32 firmware runs a state machine, packages telemetry, pushes to Oracle server. FFT + ML classifies events.
// SYSTEM ARCHITECTURE
Edge sensing → Transport telemetry → Cloud ingestion, storage, dashboard, and ML inference.
ESP32 DEVKIT V1
Central MCU + Wi-Fi
MPU6050 IMU
I2C — ±8G accel + gyro
NEO-6M GPS
UART — lat/lon/speed
LED Matrix
GPIO — status diagnostics
HTTP/JSON PUSH
Structured time-series payloads — Speed, Lat, Lon, Crash-Flag — pushed at fixed intervals. Crash-Flag payloads prioritized for immediate dispatch.
PAYLOAD STRUCTURE
{
"speed": 42.7,
"lat": 28.6139,
"lon": 77.2090,
"crash_flag": false,
"timestamp": "..."
}Telemetry Ingestion
Receive, timestamp, store
Live Dashboard
Map tracking + speed history
Crash Event Log
Timestamp, location, severity
ML Inference
FFT features → classification
// V2 HARDWARE STACK
Central MCU — runs state machine, sensor fusion, Wi-Fi client, and cloud push logic.
3-axis acceleration (±8G) + 3-axis angular velocity for impact and rollover detection.
Real-time latitude, longitude, and speed — independent of any mobile device.
Blue = network lock, Green = nominal, Red = crash/fault — instant field-level status.
Hosts backend API and dashboard for time-series telemetry ingestion and visualization.
// FIRMWARE STATE MACHINE
The firmware runs a finite state machine — the device's health is visible instantly in the field through an LED diagnostic matrix.
Wi-Fi association complete and NEO-6M has achieved satellite lock. System is initializing and acquiring GPS fix.
Telemetry (speed, GPS coordinates) is actively and successfully reaching the Oracle server. Everything is operational.
MPU6050 has detected a crash/rollover signature, or cloud connection has dropped — prompting local buffering and retry logic.
GPS ACQ
Satellite Lock
TRANSMIT
Steady State
FAULT/CRASH
Alert + Retry
// SIGNAL PROCESSING
Raw accelerometer data is noisy — simple G-force thresholds produce false positives from potholes and speed bumps. AnomX V2 applies Fast Fourier Transform (FFT) to separate genuine crash signatures from ordinary road noise.
Raw ax/ay/az accelerometer data — crash impulse is buried in road vibration noise. Hard to distinguish from potholes via amplitude alone.
After FFT: road vibration appears as narrow low-frequency peaks, while crash events show as broadband high-energy signatures — clearly separable.
Rolling window of raw ax, ay, az + gyro samples at fixed rate
Apply Hamming/Hanning function to reduce spectral leakage
Compute frequency-domain vibration spectrum per axis
Dominant freq, spectral energy, peak magnitude, band-power ratios
Compare against road vibration vs crash impact signatures
WHY FREQUENCY ANALYSIS MATTERS
A crash impulse is a short-duration, high-amplitude, broadband event, while road vibration is periodic and concentrated in narrow, predictable frequency bands. By analyzing the vibration spectrum rather than raw amplitude alone, AnomX V2 distinguishes true crash impulses from routine vibration — meaningfully reducing false-positive alerts.
// MACHINE LEARNING
FFT features feed into a supervised ML classifier that evaluates the shape and frequency content of events alongside magnitude — delivering reliable, generalizable crash classification.
Labeled runs of normal riding, braking, potholes, and simulated crash events captured from MPU6050.
FFT-derived spectral features + time-domain stats: peak-G, jerk, spectral energy, angular velocity spikes.
Supervised classifier (Decision Tree / Random Forest / lightweight neural net) trained to separate normal, harsh event, and crash classes.
Trained model runs against live feature windows to flag crash events in real-time, parallel with threshold-based triggers.
Estimated classification accuracy across detection methods
// TESTING & VALIDATION
Independent validation of MPU6050 and NEO-6M communication before integration.
Field testing of tethered connectivity and push reliability to Oracle server.
Validated FFT feature separation between vibration and genuine crash classes.
Full validation across handshake, nominal, and fault/crash conditions.
// CHALLENGES & LIMITATIONS
Relies on phone hotspot rather than dedicated cellular/SIM uplink — limits range.
NEO-6M lock time in poor sky-visibility or urban canyon conditions.
ML classifier generalization requires more diverse crash data across vehicle types.
Oracle Free Tier limits under high-frequency telemetry at scale.
// FUTURE SCOPE
Onboard SIM/GSM/LTE to remove Wi-Fi tethering dependency.
On-device ML using TinyML to reduce cloud round-trip latency.
Automated ambulance/contact dispatch from confirmed Crash-Flag.
Barometric altitude + temperature for richer rollover context.
Dual Platform
Active Features
Architecture
Crash Detection
“Built by Eurt-labs to protect those who need it most.”
Free to download. No subscriptions. No compromises.
Download AnomX FreeNo ads · No subscription · Open permissions