Good Error-Correcting Codes based on Very Sparse Matrices-英文文献.pdfVIP

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 45, NO. 2, MARCH 1999 399 Good Error-Correcting Codes Based on Very Sparse Matrices David J. C. MacKay Abstract— We study two families of error-correcting codes can be produced [1]; and an explicit algebraic construction defined in terms of very sparse matrices. “MN” (MacKay–Neal) of very good codes for certain channels was given in 1982 codes are recently invented, and “Gallager codes” were first [19]. But no practical decoding algorithm is known for any of investigated in 1962, but appear to have been largely forgotten, in spite of their excellent properties. The decoding of both codes these codes, and it is known that the general linear decoding can be tackled with a practical sum-product algorithm. problem (find the maximum-likelihood source vector in the We prove that these codes are “very good,” in that sequences of equation , where is a generator matrix, codes exist which, when optimally decoded, achieve information is a noise vector, and is the received vector) is NP- rates up to the Shannon limit. This result holds not only for complete [10]. Convolutional codes (which can be viewed the binary-symmetric channel but also for any channel with symmetric stationary ergodic noise. as block codes with memory) can approach the Shannon We give experimental results for binary-symmetric channels limit as their constraint length increases but the complexity and Gaussian channel