Decoding of Concatenated Codes for Noisy Channels With Unknown Offset

Abstract

In communication and storage systems, noise and interference are not the only
disturbances during the data transmission, sometimes the error performance is also seriously degraded by offset mismatch. We consider a simple channel such that the received signal is distorted by noise and offset mismatch, that is, r = x+v+b1, where x=(x1, x2, . . . , xn) is the transmitted codeword from a codebook, v = (v1, v2, . . . , vn) 2 Rn is the noise vector, where the vi are independently normally distributed with mean 0 and standard deviation _, b is a real number representing the channel offset, 1 is the real all-one vector (1, . . . , 1) of length n, and r 2 Rn is the received vector. Minimum modified Pearson distance (MMPD) detection has been proposed [1] as an alternative to minimum Euclidean distance (MED) detection to counter the effects of offset mismatch. A major concern, however, is the fact that the evaluation of MMPD is an exhaustive search over all candidate codewords which is infeasible for large codes. Various block codes have been proposed [2] to get good performance for channels with both noise and offset if the MMPD detection is used.