Joint erasure marking and viterbi algorithm for detecting impulsive noise corrupted signals

Abstract

In many real-world communication systems, the extent of non-Gaussian impulsive noise (IN) rather than Gaussian noise, poses practical limits on the achievable system performance. However, accurate IN statistics are typically unavailable at the receiver. We previously introduced the joint erasure marking Viterbi algorithm (JEVA), which provides significant performance improvement over the separate erasure marking and Viterbi decoding. JEVA is similar to list Viterbi algorithm (LVA) in that they both provide a list of outputs and have comparable implementation complexity. In this work, we compare the performance of the JEVA with that of the LVA with erasure inputs (ELVA). I. JOINT ERASURE MARKING AND VlTERBl DECODING In this work, we introduce the joint erasure marking Viterbi algorithm (JEVA) (l), where the erasure marking is merged with VA. Given the maximum number K of erasures to be marked, JEVA outputs a list of most likely codewords with 0, ..., k, ..., K erasures, respectively. The decoded codeword can be determined by selecting the most likely codeword with the minimum number of erasures (k_CK) that satisfies the error checking test. Based on the trellis representation of the convolutional codes, the most likely sequence with k erasures can be selected by minimizing the path metric for all possible state transition sequences and erasure pattems. The branch metric c, (i. j) associated with the transition from state i at time t-1 to statej at time t is the Euclidean distance between the received signal and the transmitted signal. The branch metrics of the k erased symbols are