基于模态分解的MEMS矢量水听器信号去噪及应用

发布时间：2018-10-23 14:54

【摘要】：矢量水听器是在标量水听器基础上发展的一种新形式,以其在定位定向方面优于标量水听器而成为研究的热门方向,再将微机电系统(MEMS)技术应用于矢量水听器更是一种创新方法和创新原理的尝试。MEMS矢量水听器具有矢量性、体积小、一致性好和可批量生产等优势。随着科学技术的不断发展,MEMS矢量水听器种类日趋繁多并且其性能也逐渐变得成熟,但其在接收信号数据时仍会混入噪声,为能更好地进行下一步的目标定向定位或是成像研究,需要先对水听器阵列信号进行去噪处理。本文系统地研究了不同的模态分解方法在MEMS矢量水听器信号去噪及应用。利用仿真信号数据和中北大学国防重点实验室在汾河进行的汾机实测数据验证了不同模态分解的去噪效果和性能指标。论文主要研究的内容包括:(1)传统的信号去噪方法,如傅里叶变换法、自适应去噪法和形态滤波法等,在应用于水声微弱信号中都有一定的去噪效果,但也存在一定的不足之处。本文利用模态分解方法对含噪信号分解的直观性和易实现性,先对仿真含噪信号进行分解处理,然后根据模态分解方法的去噪原理对分解信号进行去噪处理,得到了不同方法对仿真信号的去噪效果和性能指标,对比仿真实验的去噪效果和性能指标得出变分模态分解方法较一系列的经验模态分解方法更优。(2)由于经验模态分解方法在分解含噪信号时所产生的模态混叠影响,导致在选取固有模态函数时会产生信号失真和去噪效果差的问题,本文通过对分解后的固有模态函数进行有针对性的再去噪处理来提升算法的去噪能力。根据信号处理的基本知识,随机噪声基本都处于高频部分,将含有大部分噪声的固有模态函数直接舍去或是再利用小波阈值去噪和小波包去噪分别处理有明显信号的固有模态函数,进一步提升去噪效果和降低信号失真。(3)由于汾机实测数据的复杂性,只利用经验模态分解方法不能很好地去除实测数据中的噪声。根据对汾机实测数据的频谱分析,得到实测数据不仅有高频的随机噪声,而且还有低频的漂移干扰。结合实际问题需要,本文将经验模态分解方法与小波去噪相结合的方法应用到实测数据去噪处理中,得到了源余弦信号的较好恢复,进而在一定程度上降低了模态混叠的影响。最后,再利用变分模态分解算法对实测数据进行去噪处理,得出该方法很好地解决了经验模态分解方法在分解含噪信号时所产生的模态混叠影响,并且在去噪效果上比经验模态分解和小波结合方法更有优越性。
[Abstract]:Vector hydrophone is a new form developed on the basis of scalar hydrophone. It is a hot research direction because it is superior to scalar hydrophone in orientation and orientation. The application of MEMS (MEMS) technology to vector hydrophone is an attempt of innovation method and principle. MEMS vector hydrophone has the advantages of vector, small size, good consistency and batch production. With the development of science and technology, the MEMS vector hydrophone is becoming more and more diverse and its performance is becoming mature, but it will still mix with noise when it receives the signal data, so it can better carry out the target orientation or imaging research in the next step. First, the hydrophone array signal needs to be de-noised. In this paper, different mode decomposition methods for MEMS vector hydrophone signal denoising and their applications are systematically studied. The denoising effect and performance index of different modal decomposition are verified by using the simulated signal data and the measured data of Fen machine carried out by the National Defense key Laboratory of Zhongbei University in Fenhe River. The main contents of this paper are as follows: (1) the traditional signal denoising methods, such as Fourier transform method, adaptive denoising method and morphological filtering method, have a certain denoising effect in weak underwater acoustic signals, but there are still some shortcomings. In this paper, the method of mode decomposition is used to decompose the noisy signal directly and easily. Firstly, the simulated noisy signal is decomposed, and then the decomposed signal is de-noised according to the principle of the modal decomposition method. The de-noising effect and performance index of different methods for simulation signal are obtained. Comparing the denoising effect and performance index of the simulation experiment, it is concluded that the variational mode decomposition method is better than a series of empirical mode decomposition methods. (2) because of the modal aliasing effect of the empirical mode decomposition method when decomposing the noisy signal, The problem of signal distortion and poor denoising effect will occur when selecting the inherent mode function. In this paper, the de-noising ability of the algorithm is improved by rescinding the decomposed inherent mode function. According to the basic knowledge of signal processing, random noise is basically in the high frequency part, The inherent mode function with most noise is removed directly or the inherent mode function with obvious signal is processed by wavelet threshold denoising and wavelet packet denoising respectively. (3) because of the complexity of the measured data of Fen machine, only the empirical mode decomposition method can not remove the noise in the measured data. According to the spectrum analysis of the measured data, the measured data have not only high frequency random noise, but also low frequency drift interference. In this paper, the empirical mode decomposition method combined with wavelet de-noising method is applied to the real data denoising processing. The better recovery of source cosine signal is obtained, and the influence of modal aliasing is reduced to a certain extent. Finally, the variational mode decomposition algorithm is used to Denoise the measured data, and it is concluded that the empirical mode decomposition method can solve the modal aliasing effect when decomposing the noisy signal. And the denoising effect is better than the empirical mode decomposition and wavelet combination method.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB565.1

【参考文献】