Architecture Report
A layered system for synchronized wearable gait analysis.
The platform is organized around four connected layers: wearable acquisition, synchronized wireless transport, signal processing, and a clinician-facing dashboard. Each layer is designed to keep sessions portable, repeatable, and easy to interpret.
Wearable Nodes
Compact modules are attached to lower-body segments such as the foot, shank, thigh, pelvis, or trunk.
- 9-axis IMU for acceleration, angular velocity, and magnetic field
- Low-power microcontroller with BLE communication
- Battery and power management for complete walking sessions
Synchronization
Multi-node gait analysis depends on keeping sensor streams aligned despite wireless jitter and occasional packet loss.
- Timestamping at acquisition
- Shared frame numbering across each session
- Gateway drift correction and interpolation when needed
Signal Processing
Raw IMU samples are transformed into stable signals that can support gait event detection and biomechanical interpretation.
- Bias correction, calibration, and noise filtering
- Orientation estimation using Madgwick, Mahony, or EKF-based fusion
- Heel-strike, toe-off, stance, and swing segmentation
Clinical Interface
The dashboard prioritizes session control, live signal confidence, metric summaries, and exportable clinical documentation.
- Node connectivity and synchronization status
- Real-time cadence, timing, and symmetry indicators
- CSV/PDF-ready session outputs
Design Priorities
Portable, low-cost, and clinically practical.
The system is intended for real physiotherapy workflows, where setup time, comfort, robustness, and interpretability matter as much as raw sensor quality.
Sampling100-200 Hz target range
TransportBLE with timestamped packets
Sync target<5 ms inter-node skew
Latency target<100 ms to dashboard
Data Management
Local-first by default, exportable when needed.
Session data should be stored with minimal patient information, pseudonymized identifiers, and clear export paths for analysis or documentation.
- Acquisition: raw timestamped IMU streams per node.
- Processing: calibrated signals, orientation estimates, gait events.
- Output: session metrics, plots, CSV exports, and report summaries.