Construction of LDPC Codes Using Randomly Permutated Copies of Parity Check Matrix

Abstract

Low-density parity-check codes (LDPC) have been shown to have good error correcting performance, putting in mind the Shannon's limit approaching capability. This enables an efficient and reliable communication. However, the construction method of LDPC code can vary over a wide range of parameters such as rate, girth and length. There is a need to develop methods of constructing codes over a wide range of rates and lengths with good performance. This research studies the construction of LDPC codes in randomized and structured form. The contribution of this thesis is introducing a method called "Randomly permutated copies of parity check matrix" that uses a base parity check matrix designed by a random or structured construction method such as Gallager or QC-LDPC codes respectively to get codes with multiple lengths and same rate of the base matrix. This is done by using a seed matrix with row and column weights of one, distributed randomly and can be addressed by a number in the base matrix. This method reduces the memory space needed for storing large parity check matrices, and also reduces the probability of failing to construct a parity matrix by random approaches. Numerical results show that the proposed construction performs similarly to random codes with the same length and rate as in Gallager's and Mackay's codes. It also increases the girth average of the Tanner graph and reduces the number of 4 cycles in the resulted matrix if exists in a base graph.