Multiuser detection in OFDM-IDMA systems using noisy Independent Component Analysis

In this paper, two types of detectors, namely ordinary (noise-free) independent component analysis (ICA) and noisy-ICA are proposed. The purpose is to assist the symbols estimation and interferences suppression in orthogonal frequency division multiplexing - interleave division multiple access (OFDMIDMA) systems. The proposed detectors are implemented as post processors attached to OFDM-IDMA receiver. First the OFDMCDMA (code division multiple access) performances with ICA based blind detectors are presented. Then the proposed detectors and their applications to OFDM-IDMA system are discussed. Numerical simulations reveal how the proposed method minimizes the channel noise bias in the classical ICA algorithm, and reduces the noise through dimensional reduction.