On the Design of Near-Optimal Variable-Length Error-Correcting Codes for Large Source Alphabets

In this paper, we focus on near-optimal designs for variable-length error-correcting (VLEC) codes, while considering the performance in terms of both the average codeword length (ACL) and the symbol-error rate (SER). The criteria for narrowing the search space and enhancing the SER performance are investigated. An efficient code construction algorithm is then devised based on a random search process. Taking advantage of the significantly reduced search complexity, we are able to construct near-optimal VLEC codes for large source alphabets. Numerical results obtained for various free distance values and alphabet sizes show that constructed VLEC codes have both reduced ACL values and enhanced SER performances. Performance improvement is more notable for large source alphabets.