Deep JSCC for Visible Light Communication

PyTorch, Analog Circuit, Signal Processing — Jan. 2026

A hardware-in-the-loop Deep Joint Source-Channel Coding (Deep JSCC) system that transmits semantic features via analog light intensity using a custom 8-bit R-2R DAC.

Key Features:

  • Custom 8-bit R-2R DAC for analog light modulation
  • Autoencoder in PyTorch with differentiable channel approximation to learn robust feature representations against non-linear hardware noise
  • Achieved 98% compression on MNIST with preserved reconstruction quality
  • Custom channel linearization algorithm for optimized transmission

Technologies: PyTorch, Analog Circuit Design, Signal Processing, C++