🎵 Mood Music Player

Real-time emotion detection that plays music matching your mood

Built by Ronit Gulia

---

📖 What is this?

Mood Music Player watches your face through your webcam in real time, detects your current emotional state using a deep learning model, and automatically opens a matching YouTube playlist — all without you lifting a finger.

Point the camera at your face. The app does the rest.

---

✨ Features

• 🎯 7 emotions detected — happy, sad, angry, neutral, surprise, fear, disgust
• 🧠 OpenCV DNN face detector — works at angles, in low light, and with partial faces (far more robust than classic Haar cascades)
• 📊 EMA temporal smoothing — eliminates jittery/flickering emotion labels between frames
• 🗳️ Voting-window confirmation — an emotion is only acted on when it wins 60%+ of 8 consecutive frames, preventing false triggers
• 🎨 CLAHE preprocessing — handles backlit rooms and uneven webcam lighting gracefully
• ⏱️ Per-emotion cooldowns — surprise triggers sooner (60s), neutral waits longer (180s)
• ⚡ Instant mood-shift override — a strong emotion change (e.g. neutral → angry) bypasses the cooldown immediately
• 📝 Mood history log — all detections saved to mood_log.csv with buffered I/O
• 🔊 Optional voice announcements — announces detected mood out loud via pyttsx3

---

🛠️ Tech Stack

Layer	Technology
Language	Python 3.10+
Face Detection	OpenCV DNN (SSD ResNet, TensorFlow)
Emotion Model	FER+ ONNX (8-class, via OpenCV DNN)
GUI Framework	PyQt5
Preprocessing	CLAHE (Contrast Limited Adaptive Histogram Equalization)
Smoothing	Exponential Moving Average (EMA, α=0.35)
Confirmation	Rolling voting window (8 frames, 60% threshold)
Voice	pyttsx3 (offline TTS)
Logging	Buffered CSV writer
Threading	concurrent.futures.ThreadPoolExecutor

---

🚀 Getting Started

Prerequisites

• Python 3.10 or newer
• A working webcam
• Internet connection (first run only — downloads model files automatically)

Installation

# 1. Clone the repo
git clone https://github.com/ronitgulia/Mood_Music.git
cd Mood_Music

# 2. Install dependencies
pip install -r requirements.txt

# 3. Run
python main.py

First run: The app will automatically download two model files (~42 MB total) — the FER+ emotion model and the OpenCV DNN face detector. This only happens once. After that, startup is instant.

---

🎭 Emotions & Music

Emotion	Color	Music Vibe	Cooldown
😊 Happy	Gold	Upbeat / Bhangra	120s
😢 Sad	Blue	Calm / Melodic	150s
😠 Angry	Red	Energetic / Intense	90s
😐 Neutral	Grey	Lo-fi / Chill	180s
😲 Surprise	Pink	Dramatic / Energetic	60s
😨 Fear	Purple	Devotional / Calming	120s
🤢 Disgust	Green	Soft / Romantic	120s

---

🏗️ Project Structure

Mood_Music/
│
├── main.py             # Entry point — run this
├── gui.py              # PyQt5 main window, live camera feed, all UI widgets
├── emotion_engine.py   # Face detection + emotion inference pipeline
│                       #   └─ DNN face detector → CLAHE → FER+ ONNX → EMA → voting
├── music_manager.py    # Per-emotion cooldowns, mood-shift bypass, YouTube opener
├── mood_logger.py      # Buffered CSV mood history logger
├── config.py           # All tunable settings, emotion colours, song library
└── requirements.txt    # Python dependencies

---

⚙️ Configuration

All behaviour can be tuned in config.py:

CONFIDENCE_THRESHOLD = 40     # Min smoothed confidence (%) to accept a result
ANALYSIS_INTERVAL    = 0.35   # Seconds between analysis frames
SMOOTHING_ALPHA      = 0.35   # EMA weight (higher = more reactive)
VOTE_WINDOW          = 8      # Frames in the confirmation window
VOTE_THRESHOLD       = 0.60   # Fraction of votes needed to confirm

---

🔬 How the Detection Pipeline Works

Webcam frame (30fps)
       │
       ▼ every 0.35s
┌─────────────────────────┐
│  OpenCV DNN Face Detect │  ← SSD ResNet, detects faces at any angle
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│  CLAHE Preprocessing    │  ← Fixes uneven/backlit lighting
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│  FER+ ONNX Model        │  ← 8-class emotion probabilities
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│  EMA Smoothing (α=0.35) │  ← Removes single-frame noise/jitter
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│  Voting Window (8 frames│  ← Confirms only when 60%+ agree
│  / 60% threshold)       │
└────────────┬────────────┘
             │
             ▼
      🎵 Open YouTube

---

👨‍💻 Author

Ronit Gulia

• GitHub: @ronitgulia
• Project: Mood Music Player

---

_{Made with ❤️ and a lot of facial expressions}