Edge AI · Wearable Biosignals

A vest that thinks on the athlete

On-device health tracking for blind football players

A skin-tight sensor bib fuses seven biosignals and runs the models on-body, so athletes who can't self-monitor get real-time health insight with no phone and no cloud.

7 fused signals
< 8 ms on-device inference
500+ athletes reached

Real-time health, sensed on the body

A wearable for blind football players, where knowing an athlete's true physiological state matters more because they can't rely on visual cues. The vest fuses HR, SpO2, HRV, skin temperature, and IMU activity on-device, cleans the noise off a moving body, and runs a tiny quantized model in real time. Built with a national organization to mass-produce and distribute it.

Fig 1 — Seven raw biosignal streams fuse into one clean health readout on-device.
01

Sensor fusion

Seven channels (HR, SpO2, HRV, skin temp, IMU) fused in real time into one physiological picture.

02

Artifact rejection

A denoising autoencoder plus adaptive Kalman filter cut motion artifact 43% on a sprinting, colliding athlete.

03

TinyML on-device

A pruned, INT8-quantized model runs in under 8ms on a microcontroller, under a milliwatt budget, no cloud.

04

Built to ship

Partnered with a national organization to mass-produce and distribute to 500+ blind football players.

Fig 2 — Motion artifact rejection. Grey: raw signal on a moving body. Red: cleaned output.
7
Fused biosignal channels
43%
Motion artifact reduced
<8ms
On-device inference
500+
Athletes reached
Edge AI TinyML Sensor Fusion Model Quantization Kalman Filtering Embedded Systems Signal Processing On-Device Inference