基于独立成分分析的噪音分离研究与应用

发布时间：2018-03-18 08:24

本文选题：空调　切入点：噪音分离　出处：《华中科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着我国现代化工业水平的提高,国家经济不断向前发展,空调的使用率越来越高。在空调运转过程中,由空调部件振动而其引起的噪声蕴含着空调运行状态的重要信息,但是多个部件的振动和噪声同时出现,各类噪声互相混合在一起导致很难提取到准确有用的信息。因此,需要寻找一种比较理想的噪音分离方法,从采集到的混合噪音信号中提取出各个部件对应的声信号,从而及时有效地监测空调的状态并准确进行故障诊断。独立成分分析(independent component analysis,ICA)是近几年发展起来的信号处理方法,这种方法不依据任何先验知识,依据信号的统计特性,从观测到的混合信号中恢复出独立成分。首先简单介绍了课题研究背景和意义,以及国内外的研究现状,对独立成分分析理论基础进行了系统阐述,包括概率与统计知识,信息论知识,独立成分分析的数学模型、可分离性条件等。然后研究了独立成分分析的预处理过程:中心化过程和白化过程,实现了独立成分分析中常用的两个算法:信息最大化算法(Informax)和基于负熵最大化的快速算法(Fast ICA)。在Matlab平台上用这两个算法依次对正弦混合信号,音频混合信号,噪音混合信号开展了分离实验,从多个方面上对比分析了这两个算法在分离混合信号的有效性。最后对负熵最大化的FastICA算法在空调噪音上的应用作了进一步探讨,建立空调噪音分离的应用模型,并通过实验提取出了不同部件的噪声,获得了较好的分离效果,验证了FastICA算法在空调噪音分离方面的可行性。
[Abstract]:With the development of modern industry and the development of national economy, the utilization rate of air conditioning is getting higher and higher. In the process of air conditioning operation, the noise caused by the vibration of air conditioning components contains important information of the operating state of air conditioning. But the vibration and noise of many parts appear at the same time, and all kinds of noise are mixed together to make it difficult to extract accurate and useful information. Therefore, it is necessary to find an ideal method of noise separation. The corresponding acoustic signals of each component are extracted from the collected mixed noise signals, so that the condition of air conditioning can be monitored and the fault diagnosis can be carried out in a timely and effective manner. Independent component Analysis (ICA) is a signal processing method developed in recent years. In this method, independent components are recovered from the observed mixed signals without any prior knowledge and according to the statistical characteristics of the signals. Firstly, the background and significance of the research are briefly introduced, as well as the current research situation at home and abroad. The theoretical basis of independent component analysis (ICA) is systematically expounded, including probability and statistics knowledge, information theory knowledge, mathematical model of independent component analysis, Then we studied the pretreatment process of independent component analysis: centralization process and whitening process. Two common algorithms in Independent component Analysis (ICA) are implemented: information maximization algorithm (Informax) and Fast Matlab algorithm based on negative entropy maximization. The two algorithms are used to mix sinusoidal signal and audio signal in turn on Matlab platform. The separation experiments of noise mixing signals are carried out, and the effectiveness of the two algorithms in separating mixed signals is compared and analyzed from several aspects. Finally, the application of FastICA algorithm with maximum negative entropy to air conditioning noise is further discussed. The application model of air conditioning noise separation is established, and the noise of different parts is extracted through experiments. The better separation effect is obtained, and the feasibility of FastICA algorithm in air conditioning noise separation is verified.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TB657.2

【参考文献】