Non Linear Joint Source Channel Coding for Broadcast Channels

A novel Joint Source Channel Scheme for the two user Gaussian Broadcast channel is developed here. The scheme transmits a Bivariate Gaussian Source over a Discrete Memoryless Additive White Gaussian Noise Channel. The scheme belongs to the family of codes termed Nested Quantization. The ultimate theoretical communication limits of the Gaussian Broadcast channel are given and the different coding schemes to achieve these limits are explained. The scheme developed in this thesis is a Hybrid Digital-Analog System similar to a particular hybrid scheme that was shown to be the only scheme to achieve the theoretical limits of the Gaussian Broadcast Channel. The scheme proposed in this thesis, however, is a zero delay mapping with minimal encoding/decoding complexity. We investigate the transmission scheme for different source correlation values and show by simulation the near optimality of the proposed scheme for a wide range of correlation values. In particular, we demonstrate the scheme for independent source pairs as well as low and high correlation values. We show the connection between the space filling curves used for point to point channels and the system proposed here.