obinexus/decibelion

🔊 DΞCIBΞLION

OBINexus AI Psychoacoustic Classifier
A Constitutionally-Compliant Math-Driven Shouting Detector

👁️‍🗨️ Overview

DΞCIBΞLION is an audio intelligence module forged in the labs of OBINexus, where noise meets logic and shouting is a feature, not a bug. It mathematically analyzes human vocal input to determine emotional projection through log-scaled loudness evaluation, using a sacred constant: 85 dB.

Anything above that? It's shouting. Anything below? You're just passionate.

🧠 Features

📏 Logarithmic Amplitude Scaling – Faithful to psychoacoustic perception.
🔊 Decibel-Based Emotion Segmentation – 85 dB: The Great Divider.
🧮 MFCC-Based CNN – Because everything's better when it's filtered through 40 coefficients of raw judgment.
🧱 Torch-Powered Architecture – Low-latency yelling recognition for real-time emotional analytics.
🧰 Dataset-ready Classifier – Accepts raw .wav files like a champ.

📂 File Structure

ObinexusEmotionNet – CNN architecture for binary classification (Not Shouting / Shouting)
VocalEmotionDataset – MFCC + decibel tagging from audio files
apply_log_scale() – Applies perceptual loudness transformation using log₁₀(1 + α|x|)
visualize_mfcc() – See your screams visualized in all their MFCC glory

🚀 Usage

pip install numpy librosa torch matplotlib

from decibelion import VocalEmotionDataset, ObinexusEmotionNet

# Load and visualize
visualize_mfcc('your_emotional_outburst.wav')

# Dataset and model
dataset = VocalEmotionDataset(['your_emotional_outburst.wav'])
model = ObinexusEmotionNet()

# Predict your shouting sins
x, y = dataset[0]
output = model(x.unsqueeze(0))  # Batch dimension

📡 The Law of the Loud

By OBINexus Constitutional Mandate:

dB >= 85 → shouting
dB < 85 → tolerable emotional projection
dB == 84 → you, specifically

📜 License

OBINexus ∞ MIT License with Embedded Shout Clause™
No yelling without constitutional justification.