| 题名 | 小波在超声检测和降噪中的应用研究 |
| 作者 | 秦健华 |
| 学位类别 | 博士 |
| 答辩日期 | 2008-05-27 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 声学研究所 |
| 关键词 | 超声波 小波 波达时间 降噪 噪声训练 |
| 其他题名 | Applications of Wavelets on Ultrasound Detecting and De-noising |
| 学位专业 | 信号与信息处理 |
| 中文摘要 | 超声波的检测和降噪在工程中有着重要而广泛的应用。 小波变换是继傅立叶变换之后，数学和信息科学应用结合的又一典范，为时频分析提供了一种灵活多变的选择，也有一些其他信号处理方法所没有的特点。小波变换在信号的奇异性检测和降噪中的许多应用越来越得到人们的重视。 本文从分析小波作为信号空间的分解基以及框架理论和多尺度逼近理论入手，介绍了一些常用的小波函数。然后将小波变换融入到工程上超声波的检测和降噪的应用中，以中国科学院声学研究所东海研究站第四研究室为地下工程建设研制的两款设备的实际应用为背景，提取并分析了工程实践中泥浆和混凝土中的超声数据。提出了两个实际碰到的难题：1）信号在混凝土中的衰减和散射严重且复杂，即声传播行为复杂；2）噪声结构复杂，难以模拟。 根据实际数据的分析情况，本文提出了两个重点考虑的因素：1）信号的幅度变化较大较快；2）信号可能会被限幅。针对这两个因素，本文设计了一个包含多个幅度不同并包含限幅信号的仿真信号用于数值分析。 针对信号波达时间的检测，本文采用了Haar小波、DB4小波和复高斯小波分别对仿真信号和实际信号进行分析，给出了效果图示。对于实小波，利用不同尺度的小波系数的相关性搜索信号起始点；对于复小波，利用同尺度实、虚部小波系数相关性搜索信号起始点，并且得到了更胜实小波的效果。 针对信号降噪，本文介绍了三种通用的方法，并且重点讨论了Donoho等提出并完善的小波域阈值降噪方法。考虑到实践中很多信号之前会有一段纯噪声背景，本文提出了一种基于噪声训练的阈值选取方法，并且采用Haar小波和DB4小波对仿真信号和实际信号进行了效果分析，证明训练阈值比通用阈值更能消除机械噪声，信号失真度和信噪比也稍好。 最后本文总结了创新点，并且提出了今后进一步深入研究的方向。 |
| 英文摘要 | Ultrasonic signal detecting and de-noising play important roles in the applications of ultrasound in engineering. Wavelet is another outstanding paragon for jointing mathematics and information processing after the Fourier transform, offering a flexible choice for signal processing and having some characteristics that others do not have. More and more interest is attracted on the application of wavelets in singularity detecting and signal de-noising. In this paper, the wavelets are described as the bases or frames of the signal space and the theory of multiresolution approximation, presented by Mallat and Meyer, is discussed. Some popular wavelets are also illustrated. Later, the applications of ultrasound detecting and de-noising on engineering are introduced, with a background of two instruments offered by Shanghai Acoustics Laboratory for the construction of underground projects, the real ultrasonic data in mud and concrete are recorded and analysed. Two difficulties should be overcomed: a) the ultrasound is seriously attenuated and diffusion is complex in concrete and mud; b) the noise can not be easily simulated. Two factors are extracted for more attention: a) the volatile amplitude of the signal, and b) the amplitude could be restrained. Upon the consideration before, a simulation signal with components of variable amplitude, including one with restrained amplitude, is composed for numerical analysis. To detect the TOA, Haar, DB4 and Gaussian complex wavelets are employed to analyse the simulation signal and real signal and illustrates the results. For real wavelets, the correlation between the wavelets transform coefficients of different scales is used to detect the initial point of signal. For complex wavelets, the correlation between the real and image part of the wavelets transform coefficients is used to detect the initial point of signal and better performance is illustrated. For signal de-noising with wavelets, 3 popular methods were introduced and then the threshold method, proposed and developed by Donoho, etc., are detailedly discussed. Noticing that a lot of signals in real projects will be initiated with pure noise background signal, a threshold basing on noise training is proposed in this paper. Haar and DB4 wavelets are employed to illustrate the effect. Better performance is got with the threshold based on noise training. A summary of innovations in this study and the future work is attached |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 71 |
| 内容类型 | 学位论文 |
| 源URL | [http://159.226.59.140/handle/311008/436]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 秦健华. 小波在超声检测和降噪中的应用研究[D]. 声学研究所. 中国科学院声学研究所. 2008. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论