SNUZZ: ANDROID-BASED OFFICE’S NOISE DETECTION APPLICATION.

Snuzz is not just a decibel meter — it’s a context-aware noise guardian designed for classrooms, offices, and libraries.

Unlike standard noise-monitoring apps that trigger alerts purely based on volume, Snuzz uses an embedded TensorFlow Lite (YAMNet) model to understand what the sound is.
It ignores safe noises (like air conditioners or typing) but instantly alerts you to critical events (such as screaming or glass breaking), even when the volume is lower than the set threshold.

---

🔴 Live Monitoring Features

• Real-time noise monitoring
• Context-aware Info Card details
• Historical logs & visual graphs

---

🧠 Smart “Traffic Light” Logic

At the core of SnuzzNoise is an intelligent decision engine that processes audio in real time using a 3-stage filtering system:

Zone	Sound Type	Logic Applied
🟢 GREEN (Safe)	Fan, Rain, Typing, AC, Heartbeat	Ignore Mode — Threshold raised to 95 dB. Prevents false alarms from machinery or ambient noise.
🟡 YELLOW (Normal)	Talking, Laughter, Music, Footsteps	Standard Mode — Uses user-defined threshold (e.g., 60 dB). Sound must persist for 5 seconds to trigger an alert.
🔴 RED (Critical)	Screaming, Glass Breaking, Baby Crying	Panic Mode — Threshold drops to 50 dB. Triggers an alert within 1 second, prioritizing safety over settings.

---

✨ Key Features

• Real-Time Decibel Metering
  Uses AudioSource.UNPROCESSED to bypass Android’s automatic gain control for raw, accurate readings.

• AI Sound Classification
  Identifies over 500+ sound types locally on-device using TensorFlow Lite (YAMNet).

• Dynamic UI Feedback
  Pulsing interface reacts to sound intensity.
  Info Card icons change based on sound source (e.g., ❤️ Heartbeat, ❄️ Air Conditioner).

• Background Monitoring
  Runs as a Foreground Service, continuing detection even when the phone is locked.

• Smart Audio Alerts

  • Uses SoundPool for zero-latency alerts
  • Includes a “Deafening Loop” prevention system that pauses listening while alert sounds play

• Historical Data Logging
  Noise events are logged with timestamps and sound classifications for later review.

---

🛠 Tech Stack

Language: Kotlin
UI: Jetpack Compose (Material 3)
Architecture: MVVM + Clean Architecture
Dependency Injection: Hilt

AI / ML

• TensorFlow Lite (YAMNet model)

Concurrency

• Kotlin Coroutines
• Kotlin Flow

Local Data

• Room Database (noise history)
• DataStore (user settings)

Audio Engine

• AudioRecord — raw PCM audio capture
• SoundPool — low-latency alert playback
• MediaMetadataRetriever — dynamic audio duration handling

---

🚀 Installation

git clone https://github.com/jsnowify/Snuzz.git