Turbo均衡中的符号方差反馈均衡器

为了降低Turbo均衡中均衡器的复杂度，该文提出了符号方差反馈均衡算法(SVFE)。该算法是对精确的线性最小均方误差估计值(LMMSE)进行Taylor展开得到的。在该算法中，先利用时不变均衡器得到初步符号估计值，再根据先验符号方差对估计值加权，最后进行时不变滤波得到更佳的符号估计值。由于用到了时变的先验符号方差信息，其性能更接近精确的LMMSE均衡器。将所提算法用于Proakis C信道下的Turbo均衡处理，和时不变均衡算法进行仿真对比，所提算法将信噪比损失从0.83 dB降到了0.17 dB，并且仍可通过快速傅里叶变换降低为对数复杂度。

Symbol-variance Feedback Equalizer for Turbo Equalization

A novel Symbol-Variance Feedback Equalizer (SVEF) algorithm is proposed to reduce the computational complexity of the equalizer in Turbo equalization. The derivation of the algorithm is based on the Taylor expansion of the Linear Minimum Mean Squared Error (LMMSE) estimation function. In the proposed scheme, the initial estimates are obtained from the time-invariant equalizer, then the estimates are weighted by the a priori symbol variances and finally filtered by a time-invariant filter to obtain better estimates. As the time-variant a priori symbol variances are utilized, the performance of the proposed equalizer is much closer to that of the exact MMSE linear equalizer. Simulation results show that the Signal-to-Noise Ratio (SNR) loss of the proposed scheme in Proakis C channel is reduced to 0.17 dB from 0.83 dB compared to the various time-invariant MMSE Turbo equalization, and its computational complexity can be reduced to logarithmical order by implementation based on the fast Fourier transform.