Skip to content

This project focuses on leveraging deep learning techniques to enhance semantic communication in wireless networks, ensuring efficient transmission by prioritising meaning over raw data.

License

Notifications You must be signed in to change notification settings

cahthuranag/SemanticCom

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Semantic Communication System Simulation Model (AI-RAN)

Introduction

This simulation model demonstrates semantic-aware transmission in wireless networks using a deep learning architecture. The system employs jointly trained CNN-based transmitters and receivers to achieve end-to-end optimization of semantic communication.

System Components

1. Transmitter Design

  • Utilises multiple CNN layers for semantic feature extraction.
  • Implements Generalized Divisive Normalization (GDN) after convolutional layers.
  • Processes input data to preserve essential semantic information while removing redundancy.
  • Final layers map extracted features to channel symbols for efficient transmission.

2. Wireless Channel Model

  • Simulates AWGN and Rayleigh fading channels.
  • Incorporates path loss and multi-path effects.
  • Features adjustable SNR levels and interference patterns to test system performance.

3. Receiver Architecture

  • Utilises CNN layers for signal processing.
  • Focuses on preserving semantic meaning rather than exact signal reproduction.
  • Achieves more efficient communication compared to traditional bit-level accuracy systems.

Training Implementation

  • Trained using the CIFAR-10 dataset.
  • End-to-end optimization for balancing semantic accuracy and reconstruction quality.
  • The loss function ensures semantic fidelity while adapting to channel conditions.

Simulation Results

  • PSNR Comparison: The received image quality is evaluated under deep learning-based wireless communication (joint source-channel coding) versus traditional models (LDPC for channel coding, BPG for source coding) at different SNR levels.
  • Image Classification Accuracy: The system's ability to classify images at the destination is measured under different SNR conditions.

About

This project focuses on leveraging deep learning techniques to enhance semantic communication in wireless networks, ensuring efficient transmission by prioritising meaning over raw data.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages