Abstract

Wireless communications use different orthogonal multiple access techniques to access a radio spectrum. The need for the bandwidth efficiency and data rate enhancing increase with the tremendous growth in the number of mobile users. One promising solution to increase the data rate without increasing the bandwidth is non-orthogonal multiple access channel. For the noiseless channel like the data network, the non-orthogonal multiple access channel is named: Binary Erasure Multiple Access Channel (BEMAC). To achieve two corner points on the boundary region of the BEMAC, a half rate code is needed. One practical code which has good performance over the BEMAC is the Low Density Parity Check (LDPC) codes. The LDPC codes receive a lot of attention nowadays, due to the good performance and low decoding complexity. However, there is a tradeoff between the performance and the decoding complexity of the LDPC codes. In addition, the LDPC encoding complexity is a problem, because an LDPC code is defined with its parity check matrix which is sparse and random and lacks of structure.
This thesis consists of two main parts. In the first part, we propose a new practical method to construct an irregular half LDPC code which has low encoding complexity. The constructed code supposed to have a good performance and low encoding complexity. To have a low encoding complexity, the parity check matrix of the code must have lower triangular shape. By implementing the encoder and the decoder, the performance of the code can be also evaluated. Due to the short cycles in the code and finite length of the code the actual rate of the code is degraded. To improve the actual rate of the code, the guessing algorithm is applied if the Belief Propagation is stuck. The actual rate of the code increases from 0.418 to0.44. The decoding complexity is not considered when the code is constructed.
Next in the second part, a regular LDPC code is constructed which has low decoding complexity. The code is generated based on the Gallager method. We present a new method to improve the performance of an existing regular LDPC code. The proposed method does not add a high complexity to the decoder. The method uses a combination of three algorithms: 1- Standard Belief Propagation 2- Generalized tree-expected propagation 3- Guessing algorithm. The guessing algorithm is impractical when the number of guesses increases. Because the number of possibilities increases exponentially with increasing the number of guesses. A new guessing algorithm is proposed in this thesis. The new guessing algorithm reduces the number of possibilities by guessing on the variable nodes which are connected to a set of check nodes. The actual rate of the code increases from 0.41 to 0.43 after applying the proposed method and considering the number of possibilities equal to two in the new guessing algorithm.