The Platform

Six sensors.
One unified intelligence.

A shoe-mounted array capturing the full biomechanical signature of human motion — continuously, passively, and at clinical-grade accuracy.

How It Works

Continuous monitoring, structured for clinical use.

GAIA is a passive, shoe-mounted device. Patients wear it during normal daily activity. The ATmega328PB MCU polls all six sensors on a deterministic 1.25-second cycle and transmits a 10-field CSV packet over BLE to the clinician's mobile app.

Sensing

Passive, in-environment monitoring

Transmits 10-field CSV packets at 0.8 Hz via Nordic UART BLE — capturing heel strike, toe-off, and swing phase data throughout natural daily activity. No lab visit. No patient effort.

Accuracy

Multi-sensor biomechanical fusion

LSM6DSO at ±2g / ±250 dps, fused with Spectra flex sensors and Interlink FSRs. Bare metal C firmware ensures deterministic, synchronized sampling across all six channels.

Clinical output

Structured gait reports

The Flutter mobile app parses GAIA's stream into gait cycle segmentation, real-time charts, and exportable CSV — shaped for clinical decision-making, not raw sensor inspection.

Sensor Array — Top-Down View
TOE ARCH HEEL FSR 1 HEEL FSR 2 TOE LSM6DSO 6-Axis IMU I²C FLEX 1 dorsiflex FLEX 2 plantarflex
Interlink FSR-402 ×2 — heel & toe pressure
STMicro LSM6DSO — ±2g accel · ±250 dps gyro
Spectra Symbol Flex ×2 — ankle kinematics
Firmware Architecture

Deterministic. Four-phase.

Bare metal C on ATmega328PB — no RTOS, no uncertainty. 1.25s cycle time, 32KB flash (46% utilized).

Phase 01

ADC Sample

Read FSR1 (PC2), FSR2 (PC3), FLEX1 (PC0), FLEX2 (PC1) via 10-bit AVR ADC on 3.3V rail.

Phase 02

I²C IMU Read

Poll LSM6DSO via software bit-bang I²C (PC4/PC5). Read 12 bytes of raw accel/gyro data.

Phase 03

CSV Format

Scale raw values to milli-g and milli-dps. Compose 10-field CSV string into 96-byte buffer.

Phase 04

BLE TX

Send packet via SDEP over SPI to Bluefruit (Nordic NUS). 1,000ms post-TX delay. Repeat.

Technical Stack

Rigorous engineering. Real clinical constraints.

Every component chosen for accuracy, reliability, and clinical-grade data integrity.

Firmware
ATmega328PB
Bare metal C — four-phase sequential model: ADC sample → I²C IMU read → CSV format → BLE TX. 32KB flash (46% utilized), 2KB SRAM. No RTOS overhead; deterministic 1.25s cycle time.
IMU
STMicro LSM6DSO
±2g accelerometer (16,384 LSB/g) + ±250 dps gyroscope (131 LSB/dps). Software bit-bang I²C on PC4/PC5 for bus isolation from SPI. Outputs scaled to milli-g and milli-dps.
Wireless
Bluefruit LE SPI Friend
Nordic nRF51822 via SDEP protocol over SPI. Nordic UART Service (NUS) for transparent serial transmission. 10-field CSV packets in 96-byte buffer at 0.8 Hz. 1,000ms post-TX delay for packet integrity.
Pressure
Interlink FSR-402 ×2
Heel (PC2) + toe (PC3) plantar pressure. Inverse response — ADC 950–1000 = no contact; ADC 50–200 = max force (>10 kg). 10kΩ voltage divider on 3.3V rail.
Kinematics
Spectra Symbol Flex ×2
Dorsiflexion (PC0) + plantarflexion (PC1). 4.7kΩ pullup for 0–1023 ADC range. ~0° flat = 600–650 counts; ~90° bend = 0–100 counts.
Mobile
Flutter App
Dart/Flutter cross-platform (iOS + Android). flutter_blue_plus for Nordic NUS stream parsing, fl_chart for real-time visualization. Gait cycle detection, CSV export to device storage.

Run a pilot at your clinic.

We're seeking physical therapy clinics, neurology practices, and rehabilitation centers for paid pilot programs.

Apply for a Pilot View Clinical Applications