Pey-Man

Turn your workout into a Pac-Man game. Pey-Man tracks fitness from MATLAB Mobile sensors and gamifies your progress in a retro Pac-Man pixel UI.

Built for MathWorks Hackathon 2025.

Demo

What It Does

• Activity classification — sit / walk / run from accelerometer windows, Bagged Trees ensemble with toolbox-free fallback
• Fatigue Index — rolling activity-load timeline with annotated peaks
• Workout Quality Score + Confidence Index — per-session composite score with model-confidence overlay
• GPS route + Speed/Altitude — geoscatter map, haversine distance, speed & altitude vs. time
• Pac-Man pixel-art UI — 100% MATLAB uifigure (no external assets), neon-arcade theme

Quick Start

% After cloning the repo, from MATLAB:
cd Pey-Man
runPeyManPixelApp          % opens the UI on bundled sample metrics

% Train + classify on a real MATLAB Mobile session:
cd source/pey_man
runLocalDataSession        % expects a .mat under ../../local_data/

Bundled sample metrics live under outputs/<session>/latest_metrics.json and are loaded automatically by runPeyManPixelApp when no session is passed.

Optional live stream path for MATLAB Online rehearsal:

cd Pey-Man
runPeyManLiveStream

Live streaming is opt-in. The judged demo path remains main, runSyntheticFatigueDemo, and file-based .mat replay. The live UI includes a task dashboard where an operator can enter a target activity, minutes, calories, and steps; Pac-Man advances when the live stream meets the task and falls back if the task is ended incomplete.

MATLAB Online (one-click)

https://matlab.mathworks.com/open/github/v1?repo=YURDAKULOGLU/Pey-Man&file=source/pey_man/main.m

Results

• Activity classifier validation accuracy printed by main.m at run-time
• IRL session evidence: docs/testing/IRL_TEST_LOG.md
• Model + data contract: docs/technical/UI_METRICS_CONTRACT.md

Repo Layout

• source/pey_man/ — MATLAB sources (pipeline, classifier, UI, exporter)
• source/matlab-mobile-fitness-tracker-master/ — starter assets from MathWorks
• docs/technical/ — algorithm, architecture, data contract, sensor notes
• docs/testing/ — IRL test runbook + log, test plan
• docs/project/ — roadmap, demo runbook, goals
• docs/tr/ — Turkish UI concept & feature notes
• outputs/ — generated per-session JSON + CSV (gitignored except samples)
• local_data/ — operator-owned .mat recordings (gitignored)
• tools/ — repo hygiene checks

Tech Stack

MATLAB R2025b, MATLAB Mobile, MATLAB Online. Statistics & Machine Learning Toolbox preferred; classifier ships with a toolbox-free centroid/rule fallback so the demo runs on a stock install.

Team

• YURDAKULOGLU — model, backend pipeline, repo
• Mertrenlab — UI design & polish
• azadbulut — data collection & presentation

Acknowledgments

• Starter: mathworks/matlab-mobile-fitness-tracker
• Inspiration: cheejinteoh/matlabhackathon (prior winner)

License

MIT — see LICENSE.