量子随机滤波器及其参数计算研究

发布时间：2019-01-27 22:47

【摘要】：在随机滤波领域中,观测信号的状态转换模型通常是未知的,设计一个不依赖于观测信号及干扰噪声先验特性的智能滤波器,对抗随机干扰具有重要作用。量子滤波器利用薛定谔方程作为万能状态转换方程,从而实现了无模型滤波。因此,改进量子滤波算法和提高其适用性具有重要的应用价值。本文为了提高量子滤波算法的准确性、稳定性以及适用性主要进行了如下的研究:(1)为了提高量子滤波器的准确性与稳定性,提出了采用可变方差的高斯核函数对观测信号进行预处理,对该方法提高滤波器稳定性和准确性的原理进行了详细分析。最后将改进的量子滤波算法与递归最小二乘滤波算法进行仿真对比,说明了改进量子滤波算法的准确性、自适应性以及灵活性。(2)为了能快速获得量子滤波器的滤波参数,通过分析量子滤波器的认知过程,将滤波参数分为与观测信号无关的系统参数和仅与观测信号相关非系统参数,并建立了非系统参数和正弦输入信号的频率及输入信噪比之间的函数关系。最后将利用本文所述方法求得的滤波参数的滤波器和利用遗传算法得到的滤波参数的滤波器进行仿真对比,说明了本文所述方法的有效性和高效性。(3)在过滤非正弦输入时,为了保证滤波的准确性,本文采用短时傅里叶变换近似在线估计输入信号的频率及输入信噪比,从而实现非系统参数的在线更新。最后通过固定非系统参数的滤波器和在线更新非系统参数的滤波器的仿真对比,说明本文所述非系统参数在线更新算法的优越性及其不足。(4)当观测信号为矢量序列时,为了充分利用其空间相关性,本文首先设计了时间复杂度较低的无反馈量子滤波算法,然后将其拓展为二维量子滤波算法。最后将二维量子滤波器和独立一维量子滤波器组进行仿真对比,反映二维量子滤波算法的优缺点。
[Abstract]:In the field of random filtering, the state transition model of observation signal is usually unknown. It is very important to design an intelligent filter which does not depend on the prior characteristics of observation signal and interference noise. The quantum filter uses Schrodinger equation as the universal state transformation equation to realize modelless filtering. Therefore, improving the quantum filtering algorithm and improving its applicability have important application value. In order to improve the accuracy, stability and applicability of the quantum filter algorithm, this paper mainly studies the following: (1) in order to improve the accuracy and stability of the quantum filter, A variable variance Gao Si kernel function is proposed to preprocess the observed signal, and the principle of improving the stability and accuracy of the filter is analyzed in detail. Finally, the improved quantum filter algorithm is compared with the recursive least square filter algorithm, and the accuracy, adaptability and flexibility of the improved quantum filter algorithm are illustrated. (2) in order to obtain the filter parameters of the quantum filter quickly, By analyzing the cognitive process of the quantum filter, the filter parameters are divided into the system parameters independent of the observed signal and the non-system parameters related only to the observed signal. The functional relationship between the non-system parameters, the frequency of sinusoidal input signal and the input signal-to-noise ratio is established. Finally, the filter with the filter parameters obtained by the method described in this paper is simulated and compared with the filter parameters obtained by genetic algorithm. The effectiveness and efficiency of the proposed method are illustrated. (3) in order to ensure the accuracy of the filtering, the short time Fourier transform (STFT) is used to estimate the frequency and the input signal to noise ratio (SNR) of the input signal. In order to realize the online update of non-system parameters. Finally, the advantages and disadvantages of the on-line updating algorithm for the non-system parameters are illustrated by the comparison between the filter with fixed non-system parameters and the filter with on-line updating parameters. (4) when the observed signal is a vector sequence, In order to make full use of its spatial correlation, this paper first designs a non-feedback quantum filtering algorithm with low time complexity, and then extends it to two-dimensional quantum filtering algorithm. Finally, the two-dimension quantum filter and the independent one-dimensional quantum filter bank are simulated and compared to reflect the advantages and disadvantages of the two-dimensional quantum filter algorithm.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN713

【参考文献】