Home Screen
Select Workout
Do workout in proper posture to score

PulseFit Arena

Inspiration

Traditional workouts often feel repetitive and boring, especially for students and young users who spend most of their time around games, music, and digital entertainment. We wanted to explore a simple idea:

“What if working out felt like playing an arcade game?”

That idea became PulseFit Arena — a rhythm-based AI fitness game where players perform squats in sync with music beats using real-time computer vision.

Inspired by games like Guitar Hero and Just Dance, we combined:

rhythm gaming
AI pose tracking
fitness
multiplayer competition
immersive visuals

into one interactive experience.

Our goal was to make workouts:

fun
competitive
immersive
social
visually exciting

instead of repetitive and boring.

What it does

PulseFit Arena is an AI-powered rhythm fitness game that transforms workouts into an interactive gaming experience using real-time pose detection and beat synchronization.

Players squat to the rhythm of the music while the system:

tracks body posture
validates squat depth
checks timing accuracy
calculates scores
provides AI-powered feedback

The gameplay combines:

computer vision
rhythm mechanics
multiplayer competition
live scoring
cyberpunk arcade visuals

into one immersive experience.

Core Features

Real-time squat detection using MediaPipe Pose
Beat-synced rhythm gameplay
Accuracy-based scoring system
AI posture correction
Multiplayer battle mode
Combo streaks and score multipliers
Cyberpunk-inspired neon UI
Post-game fitness analytics

The experience feels like:

“Guitar Hero meets AI fitness training.”

System Architecture

Webcam Feed
    ↓
OpenCV Frame Capture
    ↓
MediaPipe Pose Detection
    ↓
Squat Validation System
(Knee Angle + Depth Tracking)
    ↓
Rhythm Engine
(BPM + Beat Synchronization)
    ↓
Scoring System
(Perfect / Good / Miss)
    ↓
Game UI + Visual Effects
    ↓
AI Coach Feedback

How we built it

We built PulseFit Arena using a combination of:

game development
computer vision
audio processing
real-time gameplay systems
pose estimation AI

Tech Stack

Technology	Purpose
Python	Core application logic
Pygame	Game engine and rendering
OpenCV	Webcam capture and processing
MediaPipe Pose	Real-time body tracking
librosa	BPM and beat detection
pygame.mixer	Music playback and synchronization

Real-Time Pose Detection

We used MediaPipe Pose to detect:

hips
knees
ankles
shoulders

The system continuously calculates knee angles to determine:

squat depth
movement timing
squat completion

This allows accurate real-time movement tracking directly through a webcam.

Rhythm Engine

We built a rhythm synchronization system using:

BPM analysis
beat timestamp extraction
timing windows

This allows gameplay to function similarly to rhythm games where players must squat exactly on beat.

Scoring System

Players receive:

Perfect
Good
Miss

based on:

squat timing
movement accuracy
squat depth consistency

Combo streaks increase score multipliers and create competitive gameplay.

AI Coach

The AI coach provides live feedback such as:

“Go lower”
“Straighten your back”
“Perfect squat”
“Too early”

After each match, the system generates:

posture analytics
timing accuracy
squat count
estimated calories burned

Gameplay Pipeline

Music Beat
    ↓
Incoming Rhythm Note
    ↓
Player Performs Squat
    ↓
Pose Detection
    ↓
Timing Validation
    ↓
Score Calculation
    ↓
Combo + Visual Effects
    ↓
AI Feedback + Analytics

Challenges we ran into

One of the biggest challenges was synchronizing:

webcam input
computer vision processing
music timing
gameplay rendering

all in real time with minimal latency.

1. Accurate Squat Detection

Detecting proper squat movements consistently was difficult because:

camera angles varied
lighting conditions changed
users had different body proportions

We solved this using:

knee angle thresholds
movement smoothing
pose confidence filtering
repeated frame validation

2. Beat Synchronization

Even tiny delays between:

music playback
pose detection
scoring

made gameplay feel inaccurate.

We optimized:

audio timing
beat spawning
frame synchronization
scoring windows

to improve responsiveness.

3. Performance Optimization

Running:

computer vision
pose estimation
real-time rendering
audio systems

simultaneously initially caused FPS drops.

We improved performance by:

lowering webcam resolution
optimizing MediaPipe processing
reducing rendering overhead
separating gameplay systems logically

Accomplishments that we're proud of

We are proud that we transformed a simple fitness exercise into a fully interactive rhythm gaming experience.

Highlights

Successfully implemented real-time AI squat tracking
Built a fully playable rhythm fitness game
Created a multiplayer battle experience
Designed a polished cyberpunk-inspired interface
Achieved low-latency gameplay
Combined AI, gaming, and fitness into one experience

Most importantly, we created something that encouraged people to:

compete
move
exercise
interact socially

while genuinely enjoying the experience.

Performance Flow

Camera Input
    ↓
30 FPS Processing
    ↓
Optimized Pose Tracking
    ↓
Real-Time Beat Matching
    ↓
Smooth Gameplay Rendering
    ↓
Low-Latency Feedback