Voice Command Classifier (from Scratch)

A minimalist Keyword Spotting (KWS) system built entirely in NumPy. This project demonstrates the mathematical foundations of neural networks by implementing forward and backward propagation without high-level libraries like PyTorch or TensorFlow.

🎯 Project Goal

To build a robust classifier for "Yes", "No", and "Unknown" commands using the Google Speech Commands Dataset, focusing on efficient feature extraction and raw matrix operations.

🧠 Key Features

• Zero-Framework Inference: All neural layers and training logic are written in pure NumPy.
• Signal Processing: Audio features are extracted using Mel-frequency cepstral coefficients (MFCC) via librosa.
• Validation Pipeline: Includes a synthetic "smoke test" to verify model convergence on wave patterns vs. white noise.

🛠 Tech Stack

• Python 3.x
• NumPy (Linear Algebra & Model Logic)
• Librosa (Digital Signal Processing)

🚀 Getting Started

# Install dependencies
pip install numpy librosa

# Run the synthetic convergence test
python main.py