Distance Enumerator Analysis for Interleave-Division Multi-User Codes

Abstract

Consider an interleave-division multi-user coding model for a Gaussian multiple-access channel (MAC). Each user’s message is encoded by a separate channel encoder followed by a user-interleaver, which is employed for user separation. All the users’ codewords are jointly regarded as one multi-user codeword, and the distance between two multi-user codewords is defined as the Euclidean distance after the inter-user superposition caused by the MAC. A distance enumerator analysis estimates the multi-user maximum likelihood decoding performance. By introducing user-scramblers, a multi-user code ensemble is defined, and its distance enumerator is calculated as concatenated multiple single-user codes and a superposition code. Both finite-length and asymptotic analyses are given and it is shown that interleave-division multi-user codes have much better distance properties than conventional random direct-sequence code-division multiple-access. In fact, the codes achieve almost the same asymptotic uniquely decodable performance as a multi-user random code.