UUV阵列自适应噪声抵消关键技术研究

发布时间：2018-04-16 01:32

本文选题：UUV + 阵列自适应噪声抵消　；参考：《西北工业大学》2014年博士论文

【摘要】：随着新型无人水下航行器（Unmanned Underwater Vehicle, UUV）的发展其航速将得到进一步提高，UUV声纳阵列所接收到的自噪声和混响级也随之大幅增强，同时由于水声对抗技术的不断发展和广泛应用，如何有效地抑制和消除自噪声、混响以及干扰对远程微弱目标探测的影响，提高UUV阵列对微弱信号检测的能力是目前进行海洋资源探测开发和加强海防的现实迫切需求。本文在对UUV声纳阵列自噪声产生机理和特性分析的基础上，针对阵列信号多通道处理的特点，将基于多通道差分方法与基于核函数的非线性自适应滤波器技术相结合，重点研究了基于核函数的非线性自适应滤波器理论收敛性能、非平稳环境中基于单核函数以及基于多核函数的线性与非线性加权组合的阵列自适应噪声抵消方法，针对多输入多输出（MIMO）阵列的空时两维自适应处理对混响和干扰进行抑制的方法，并通过仿真实验对本文所提方法的有效性进行了验证分析。本文的主要研究成果和创新点如下： 1.针对动态非平稳输入信号会导致基于核函数的非线性自适应滤波器所构造的在线“字典”中出现和输入信号统计分布不匹配的冗余失效元素问题，提出了存在冗余失配“字典”元素情况下基于单高斯核函数方法的最小均方误差KLMS非线性滤波器误差均值和均方收敛特性的理论计算方法，为非平稳环境中基于单核函数的非线性自适应滤波器性能分析和设计提供了有力理论工具。仿真结果表明：基于所推理论计算方法预测出的收敛曲线与蒙特卡洛仿真实验平均后所得的均方误差学习曲线在瞬时动态过程和稳态阶段均一致吻合。因此不仅验证了所推理论计算方法的正确性和有效性，而且该理论计算方法为非线性自适应噪声抵消滤波器在非平稳动态应用中针对“字典”提出自适应更新准则提供了理论依据。 2.针对动态非平稳噪声环境，在多通道差分方法提供相关参考噪声情况下，提出了具有1-范数的FOBOS-KLMS-1和自适应-范数的FOBOS-KLMS-a两种促在线“字典”稀疏自适应噪声抵消方法，同时证明了在引入1-范数促稀疏操作后，所提两种FOBOS-KLMS方法在均值意义上仍然是平稳且严格收敛的。两种FOBOS-KLMS方法通过对基于单高斯核函数的非线性自适应滤波器引入1-范数稀疏正则项后，得到以向前向后算子分裂方法定义的在线“字典”元素自适应更新策略，即对在线“字典”中对函数拟合估计贡献权值小于给定门限的“字典”元素进行删除操作。利用湖试噪声数据的仿真结果表明：在声纳阵列被加速和减速的非平稳变化过程中，与常规线性方法相比核自适应滤波方法对噪声估计的均方误差低了7dB，而且所提两种“字典”自适应稀疏方法降低了阵列非线性自适应噪声抵消器的“阶数”，因此计算复杂度和对存储空间的要求更低，为工程实际中阵列在线自适应噪声抵消应用奠定基础。 3.针对基于多核的方法较之单核方法具有更多的系统自由度和特征能够有效解决动态系统在线辨识和核函数参数必须离线选择问题的优点，提出基于K个高斯核函数的最小均方误差MKLMS1、MKLMS2、MKLMS3三种多核自适应滤波算法，并提出了前两种多核滤波器在预先给定“字典”元素情况下的理论收敛分析计算方法，通过所推理论表达式可以比较三种不同类型的多核自适应滤波器的性能特点。仿真结果表明：基于所提计算方法预测出的理论收敛曲线与蒙特卡洛仿真实验平均后所得的误差学习曲线在瞬态阶段和稳态阶段均一致吻合，，不仅验证了所推多核自适应滤波器理论性能计算方法的正确性和有效性，而且提供了基于多核函数的非线性自适应滤波器性能分析、比较和设计手段。 4.针对阵列接收噪声组成分量的空时复杂性，同时根据多核函数的自适应滤波器结构，提出两种基于线性核函数与非线性高斯核函数加权组合的双核归一化最小均方误差滤波BKNLMS1方法和BKNLMS2方法。针对阵列多通道差分方法提供相关噪声的复杂性，通过对线性核函数和非线性高斯核函数分别加权得到两种综合自适应滤波器。不仅考虑线性相关噪声的抵消，而且进行非线性相关噪声的抑制，并利用湖试噪声数据分别对单频和调频接收信号进行相关检测验证阵列自适应噪声抵消效果。仿真结果表明：所提两种方法可以同时自适应抵消线性和非线性噪声分量从而改善信噪比提高检测概率，在相同检测概率下不仅相对传统线性自适应噪声抵消器的检测概率提高了5dB，而且比单核KNLMS算法非线性自适应噪声抵消方法提高了2dB，因此所提两种方法在对声纳阵列自适应噪声抵消的工程实际中具有很强的实用价值。 5.针对UUV舷侧MIMO阵列对低速运动目标检测时易受到混响和干扰影响的问题，提出针对MIMO阵列基于子空间估计降维的空时两维自适应处理对混响和干扰进行抑制的方法。该方法结合扁长椭球波函数的时限带限特性近似构造出降维的杂波子空间，并利用与发射波形正交的辅助匹配滤波通道估计出干扰加噪声协方差矩阵，通过“逼零”方法求得MIMO阵列系统的空时权矢量。仿真结果表明：当存在非理想因素影响时，该方法与其它方法相比能够更有效抑制混响和干扰且UUV舷侧MIMO-STAP降维运算量更低。本文研究成果对改善信噪比提高UUV阵列对远程微弱信号检测性能具有重要的理论意义和实用价值，对其它水下声纳阵列系统的降噪问题解决具有借鉴意义。
[Abstract]:With the new type of unmanned underwater vehicle (Unmanned Underwater, Vehicle, UUV) its speed of development will be further improved, UUV received array sonar self noise and reverberation level has been greatly enhanced, at the same time, due to the continuous development of the underwater acoustic countermeasure technology and extensive application, how to restrain and eliminate noise, reverberation and interference the influence of distance weak target detection, improve the ability of the UUV array for weak signal detection is the detection of marine resources development and strengthen the defense of urgent needs.
Based on the UUV sonar array self noise analysis of the causes and characteristics, in the array signal processing of multi channel characteristics, based on multi channel difference method and nonlinear adaptive filter technique based on kernel function combination, focusing on the kernel function of non linear adaptive filter theory convergence based on nonstationary environments based on a single kernel function and array adaptive noise cancellation method of linear and nonlinear weighted combination of multi kernel function based on multiple input multiple output (MIMO) array of two dimensional space-time adaptive processing of reverberation and interference suppression method, and verified the validity through simulation analysis methods mentioned in this paper. The main research results and innovations are as follows:
1. according to the dynamic nonstationary input signal will lead to "the construction of online nonlinear adaptive filter based on kernel function dictionary" appears and input redundant signal does not match the statistical distribution of failure elements, puts forward the existing calculation methods of single Gauss kernel function method of the minimum mean square error KLMS nonlinear filter mean error and mean square convergence properties based on the theory of "dictionary" elements under the condition of redundant, non-stationary environment in nonlinear adaptive filter performance analysis and design based on the kernel function provides a powerful theoretical tool. The simulation results show that the convergence curve and Monte Carlo simulation experiment the average calculation method of the prediction of the theory of push the MSE learning curves are consistent well in the instantaneous dynamic process and stable stage. Therefore based on not only verified the correctness of the method and the theoretical calculation The theoretical calculation method provides a theoretical basis for nonlinear adaptive noise cancellation filter in the non-stationary dynamic applications, and proposes an adaptive update rule for dictionary.
2. according to the dynamic nonstationary noise environment in multi channel differential method to provide relevant reference noise, we proposed a 1- norm and FOBOS-KLMS-1 norm of the adaptive FOBOS-KLMS-a two promote online dictionary sparse adaptive noise cancellation method is also proved in introducing 1- norm sparse Pro operation after the two the FOBOS-KLMS method is still stable and strict convergence on average. Two kinds of FOBOS-KLMS method based on the nonlinear adaptive filter of single Gauss kernel function based on the introduction of 1- norm regularization, get online to the forward backward operator splitting method definition dictionary "element adaptive updating strategy, namely the" dictionary "in the function of online the estimation of weight less than a given threshold with" dictionary "elements removed. The simulation test results using lake noise data show that: in the sonar array Column is the acceleration and deceleration of the non-stationary process, compared with the conventional linear method of kernel adaptive filtering method for noise estimation mean square error of low 7dB, and the two "dictionary" adaptive sparse method reduces array nonlinear adaptive noise canceller "order", so the computational complexity and storage the space requirement is lower, to lay the foundation for the application of array online adaptive noise cancellation in actual engineering.
3. according to the method based on multi core compared with single degree of freedom system with kernel method and more features can effectively solve the dynamic system identification and kernel function parameter selection problem of the advantages must be offline, MKLMS2 K proposed a Gauss kernel function of the minimum mean square error MKLMS1, based on MKLMS3 three multi kernel adaptive filtering algorithm, and the calculation method of the previous two kinds of theories in the given convergence of multi-core filter elements under the condition of "dictionary", the performance characteristics of the multi core expression can push the theory of adaptive filter to compare three different types. The simulation results show that the theory of error convergence curve and Monte Carlo simulation experimental average calculation method to predict the the learning curves are consistent in the transient stage and steady stage based on not only verify the push multi-core adaptive filter theory performance calculation The validity and validity of the method are also provided, and the performance analysis, comparison and design of the nonlinear adaptive filter based on the multi kernel function are provided.
4. for the array receiving noise component space-time complexity at the same time, according to the adaptive filter structure of multi kernel function, put forward two kinds of linear kernel function and kernel function weighted combination of nonlinear Gauss dual normalized least mean square error method and BKNLMS1 filtering method based on BKNLMS2. According to the complexity of multi channel array differential method provides related noise. Two kinds of comprehensive adaptive filter by weighting of linear kernel function and kernel function. The nonlinear Gauss not only consider the linear correlation of noise cancellation, and suppression of nonlinear correlation noise, and the noise data were related to the lake trial test array adaptive noise cancellation effect on single frequency and frequency modulation signal. The simulation results show that the the two methods can also offset the linear and nonlinear adaptive noise components so as to improve the SNR. In the same measuring probability, the probability of detection under not only the relative detection probability of traditional linear adaptive noise canceller improves 5dB and KNLMS than the single core algorithm of nonlinear adaptive noise cancellation method is improved by 2dB, so the two method has great practical value in the engineering practice of sonar array adaptive noise cancellation.
5. for the UUV side of MIMO array to the low speed moving target detection by reverberation and interference problems, proposed dimensionality reduction for MIMO array subspace estimation based on the two dimensional space-time adaptive processing of reverberation and interference suppression method. This method combines prolate spheroidal wave functions with characteristics similar to construct time reduction the dimension of the clutter subspace, and by using the estimated interference plus noise covariance matrix and auxiliary waveform orthogonal matching filter channel, through the "empty weight vector zero forcing method to obtain the MIMO array system. The simulation results show that when the influence of non ideal factors, compared with other methods, this method can effectively suppress the reverberation and interference and UUV side MIMO-STAP dimensionality reduction computation is lower.
The research results in this paper are of great theoretical significance and practical value for improving the signal-to-noise ratio and improving the performance of UUV array for remote weak signal detection. It is of reference for other underwater sonar array systems to solve the problem of noise reduction.

【学位授予单位】：西北工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：U674.941;TN911.23

【参考文献】