基于盲源提取的强混响背景下LFM信号回波检测

在主动声纳进行浅海区域的沉底、掩埋目标探测中，海底混响是主要的背景干扰，其严重影响了LFM信号回波的检测。基于自回归(AR)模型的预白化算法虽能适应匹配检测，但在低信混比时检测效果不甚理想；基于分数阶傅里叶变换(FRFT)的分数域滤波算法虽能抑制混响，但随着混响增强检测性能衰减较快。本文提出一种改进的基于线性预测的盲源提取算法(LP-BSE)，通过利用待提取信号与发射信号强相关的先验知识，采用最小二乘法估计并固定对应线性预测器参数；然后通过构造最小均方预测误差准则，迭代求取解混向量以提取用于匹配检测的期望回波。仿真结果表明，该算法有着更好的匹配检测效果且更稳健。串联LP-BSE与FRFT形成联合方法，匹配检测性能得到进一步提升。 

Echo Detection of LFM Signal under Strong Reverberation Background Based on Blind Source Extraction

In the detection of submerged and buried targets in shallow sea area by active sonar, bottom reverberation is the main background interference, which seriously affects the detection of LFM signal echo. Pre-whitening algorithm based on Autoregressive (AR) model can adapt to matching detection, but the detection effect is not ideal at low signal-to-reverberation ratio. Fractional domain filtering algorithm based on fractional Fourier transform (FRFT) can suppress reverberation, but the detection performance decreases rapidly with the enhancement of reverberation. An improved blind source extraction algorithm based on linear prediction (LP-BSE) is proposed in this paper. By utilizing the prior knowledge of strong correlation between the Signal to be extracted and the transmitted signal, the parameters of the corresponding linear predictor are estimated and fixed by the least square method. Then, by constructing the minimum mean square prediction error criterion, the de-mixing vector is iteratively obtained to extract the desired echo for matching detection. The simulation results show that the algorithm has better matching detection effect and is more robust. Combining LP-BSE with FRFT to form a joint method, the matching detection effect is further improved.