基于KNN信道识别算法的无前后缀频域均衡技术

Prefix-free and Suffix-free Frequency Domain Equalization Technology Based on KNN Channel Identification Algorithm

为了提高单载波频域均衡(SC-FDE)系统的频带利用率和通信速率,提出了一种基于KNN信道识别算法的无前后缀频域均衡技术;该技术通过构建不同信道环境下带有导频的训练数据集,根据训练数据通过不同信道条件下的特征并使用KNN算法进行分类,进而确定信道类型及信道最大多径时延,通过信道最大多径时延合理确定无前后缀重叠剪切频域均衡算法的最优重叠剪切长度,进而消除多径干扰对无前后缀频域均衡数据的影响,达到最优解调性能;仿真表明:基于KNN的重叠剪切算法与文献推荐的50%重叠剪切法相比,性能基本一致,即在误码率为10^(-3)的情况下,信噪比恶化不大于0.5 dB;此算法准确确定了最优重叠长度,减少了大量无效的重叠长度,保证了多径信道下系统的误码性能,达到性能和复杂度的平衡。

In order to improve the frequency band utilization and communication rate of single-carrier frequency domain equalization(SC-FDE) system, a prefix-free and suffix-free frequency domain equalization technology based on K-Nearest-Neighbors(KNN) channel identification algorithm is proposed. This technology constructs training data sets with pilots in different channel environments, and classifies the data sets by using the KNN algorithm and the characteristics of the training data under different channel conditions, and then determines the channel type and the maximum delay of the multipath channel. The optimal overlap-cut length of the frequency-domain equalization algorithm without prefix and suffix overlap-cut is reasonably determined by the maximum delay of the multipath channel, and then the influence of multipath interference on the frequency domain equalization data without prefix and suffix is eliminated, and the optimal demodulation performance is achieved. The simulation shows that compared with the 50% overlap-cut method recommended in the paper, the KNN-based overlap-cut algorithm has basically the same performance, that is, when bit error rate is 10^(-3), signal-to-noise ratio deterioration is not more than 0.5 dB. This algorithm accurately determines the optimal overlap length, reduces a large number of invalid overlap lengths, ensures the excellent bit error performance of the system in multipath channels, and achieves a balance between performance and complexity.