#📡EchoScope

Radio Signal Classification using Deep Learning

Features

• Dual-Branch Model

  • IQ Branch: Bidirectional LSTM over time-series IQ data.
  • Spectrogram Branch: Transformer-based patch encoder over spectrogram inputs.

• Web Dashboard

  • Upload .wav files for inference.
  • Visualize I/Q data.
  • View class probabilities and predicted labels.

• Architecture

  • Transformer encoder with patch embedding for spectrogram data.
  • Bi-directional LSTM feature extractor for sequential IQ.
  • Fused logits across both modalities.

Tech Stack

Backend

• Python 3.10+
• PyTorch – Deep learning framework for model development and training
• NumPy – Numerical computing and signal data handling
• Scikit-learn – Data preprocessing and evaluation
• Scipy – Actual spectrogram generation from data
• Matplotlib – Visualizations (e.g., spectrograms, attention maps)

Frontend

• React.js – Web UI framework for dashboard interface

Middleware / Server

• Node.js – Server-side runtime for dashboard backend
• Express.js – Handles file uploads and communication with Python scripts
• Python – Generates I/Q visualizations & is used for the model pipeline.

Tools & Utilities

• Jupyter Notebook – For model prototyping and analysis

---

Screenshots