Enhanced turbo detection for SCMA based on information reliability

Sparse Code Multiple Access (SCMA) is a recently designed spreading technology in which QAM symbols are mapped to different OFDMA tones. The codebook based mapping can be regarded as a spreading encoding process, where the shaping gain helps promote spectral efficiency and enhance system performance. Based on the SCMA encoding scheme, this paper proposes a joint Multi-user detection (MUD) and channel decoding approach, applying the "turbo principles". Compared to traditional separate detection and decoding scheme, turbo detection performs better with higher gain and only moderate complexity. On the top of this, an improved method is proposed, where extrinsic information is modified before iterated. The improvement is based on the reliability of information, maintaining reliable information and diminishing the unreliable. Specifically, two schemes are proposed for parameter selection. These methods model the distribution of information as Mixture Gaussian Model (GMM) and the adaptive approach uses Expectation Maximization (EM) algorithm for parameter estimation. The proposed methods perform even better than "turbo principles" in error correction and hopefully can be applied to other turbo systems.