LDPC Codes for Non-Uniform Memoryless Sources and Unequal Energy Allocation

In this paper, we design a new energy allocation strategy for non-uniform binary memoryless sources encoded by Low-Density Parity-Check (LDPC) codes and sent over Additive White Gaussian Noise (AWGN) channels. The new approach estimates the a priori probabilities of the encoded symbols, and uses this information to allocate more energy to the transmitted symbols that occur less likely. It can be applied to systematic and non-systematic LDPC codes, improving in both cases the performance of previous LDPC based schemes using binary signaling. The decoder introduces the source non-uniformity and estimates the source symbols by applying the SPA (Sum Product Algorithm) over the factor graph describing the code.