BearFit Multi-Device IMU Fitness Sensing
A multi-device IMU workout sensing prototype for body awareness in the gym, built around what people already carry plus a lightweight ring device: phone, watch, earbuds, and ring.
Overview
BearFit is designed around a low-friction gym workflow: you walk in dressed normally, bring your iPhone, wear an Apple Watch, put on your earbuds, and add a lightweight IMU ring, then the system can start tracking whole-body movement immediately without cameras, straps, or dedicated gym hardware.
Technically, the project fuses motion streams from the phone, Apple Watch, earbuds, and left/right ring IMUs, then adds local logging, MQTT streaming for low-latency Linux delivery, a FastAPI-based inference backend, and offline movement-analysis scripts for repetition progress, completion features, pose fusion, and model training experiments.
Video
Key Work
- Built a multi-device IMU pipeline for collecting synchronized motion streams from phone, watch, earbuds, and rings during workouts.
- Extended the sensing setup with ring_left and ring_right inputs to capture finer hand and arm motion during exercise.
- Added MQTT-based live streaming so multiple Linux receivers can subscribe to the same sensor session with low latency.
- Implemented a FastAPI backend for ingest, logging, exercise inference experiments, and body-movement analysis workflows.
- Created offline tooling for progress baselines, completion features, pose fusion, and visualization across multi-device IMU sessions.
