EEG Signal Processing with DGCNN & ONNX
Feb 1, 2025
·
1 min read
A deep learning-based EEG signal processing project that classifies emotional states using the DREAMER dataset. The model is built using the DGCNN architecture from TorchEEG and optimized for fast inference via ONNX.
Key Features:
- 🧠 Utilizes the DREAMER dataset for EEG-based emotion classification
- 🧩 Employs DGCNN (Deep Graph Convolutional Neural Network) tailored for EEG signal learning
- 🧼 EEG signals transformed into a 9x9 electrode grid format using TorchEEG’s
ToGrid()transform - ⚙️ Processes four frequency bands (delta, theta, alpha, beta) across 81 electrodes
- 🏷️ Supports binary classification (e.g., positive vs. negative emotion)
- 🚀 Model exported to ONNX format for lightweight and high-speed deployment
- 🔄 Data loading and preprocessing handled with PyTorch DataLoader and TorchEEG’s built-in utilities
