中文摘要：

MIMO-OFDM（多输入多输出—正交频分复用）系统的信道估计问题是系统接收机进行相干解调的关键.针对标准粒子滤波算法在MIMO-OFDM信道估计时存在观测系数不准确和粒子退化等问题,提出了一种基于神经网络的重要性样本调整粒子滤波（NNISA-PF）算法.对MIMO-OFDM通信系统及时变信道进行建模,得到了状态空间模型;详细分析了标准粒子滤波的改进算法——NNISA-PF;在MIMO-OFDM快衰落和慢衰落信道下,对Kalman、Bootstrap和NNISA-PF三种滤波算法分别在AWGN、Middleton-A噪声下的NMSE和BER性能进行了仿真对比分析.仿真结果表明,在快衰落和慢衰落情况下,NNISA-PF算法都可以有效对抗噪声干扰,尤其是在非高斯噪声环境下优势明显;NNISA-PF算法可以较准确地对MIMO-OFDM时变信道进行半盲估计,使MIMO-OFDM技术优势得到充分发挥.与现有的半盲信道估计方法相比,该方法具有估计精度高、对非高斯噪声鲁棒性强等优点.

英文摘要：

The channel estimation of MIMO-OFDM（ multiple input multiple output-orthogonal frequency division multiplexing）system is key for coher- ent demodulation. In view of the standard particle filter algorithm problems observed inaccurate coefficients and particle degradation in the channel estimation of MIMO-OFDM, an improved algorithm is proposed that is NNISA-PF. First of all, the MIMO-OFDM system and its time varying channel model are built to get the state space model. Then, the standard particle filter algorithm and its improved algorithm--NNISA-PF are introduced. Finally, in the MIMO-OFDM fast and slow fading channels, the NMSE and BER performance of Kalman, Bootstrap and NNISA-PF filtering algorithm in the AWGN and Middleton-A noise are simulated and analyzed. The simulation results show that NNISA-PF algorithm can be effectively against noise interference no matter in the fast fading or slow fading channel, especially in non Gauss noise environment. It also indicates that NNISA-PF algorithm can get semi- blind estimation of time-varying MIMO-OFDM channel accurately so that it gives full play to the advantages of MIMO-OFDM technology. Compared with the existing semi-blind channel estimation method, the method has high estimation accuracy and the advantages of non Gauss noise robustness.