次声信号特征提取与分类识别研究

发布时间：2018-07-05 13:17

本文选题：次声信号 + 特征提取　；参考：《中国地质大学(北京)》2015年硕士论文

【摘要】：次声是一种人耳听不到的低频信号,它的频率范围在0.01~20Hz之间。自然界中海啸、火山喷发、极光、地震、泥石流,人类活动中的核爆炸、火箭发射、炮兵射击等都会产生次声信号。由于各种事件激发次声的机理不尽相同,各类型事件产生的次声信号在频率轴上的能量分布不同。从监测到的次声信号本身的特点,可以反推出产生次声信号的事件类型,从而达到次声信号分类识别的目的。对次声信号的特征提取和分类识别研究一直是次声信号处理领域中的热点内容。总结前人的研究成果发现,次声信号的分类识别算法主要是沿着两条主线路不断的取得突破和发展。一方面是次声信号特征提取技术的研究;另一方面是模式识别算法的设计研究。次声事件识别的关键环节是前者,即如何从信号中提取有效的特征作为识别依据。而识别效果的好坏本质上也是由所选用的模式特征决定的。对特征提取这一方向的研究重点主要在于挖掘能够表现信号类别的特征,以及如何有效的提取这些特征的信号处理技术。分类识别这一部分的研究重点在于合适的匹配提取到的特征向量,研究各种分类模型的算法和结构,设计准确高效的分类器,完成准确划分信号类别的最终目的。本文将研究通过次声信号对自然灾害进行事件分类,研究的重点是次声信号特征提取的技术方法和模式识别分类算法的设计。本论文的主要内容如下:论文首先详细地介绍了自然灾害中次声信号分类识别的研究背景和重要意义,着重对比分析了各种次声信号特征提取算法的优点和不足;研究了三种技术方法用于次声信号的特征提取和两种分类算法用于分类识别;在此基础上,使用采集的次声数据对完整的分类模型进行试验验证。分析试验结果,对比各个方法的有效性。本论文以自然灾害中各种事件产生的次声信号为研究对象,以降低实际次声监测中的误报率为目的,研究了地震、海啸、火山、泥石流等所产生的次声信号的特征提取技术和分类识别算法。期望本研究的成果可以对次声信号处理方法和实际应用产生一定的参考作用。
[Abstract]:Infrasound is a kind of low frequency signal which can not be heard by human ear. Its frequency range is between 0.01 Hz and 20 Hz. In nature, tsunamis, volcanic eruptions, auroras, earthquakes, mudslides, nuclear explosions in human activities, rocket launches and artillery fire all produce infrasound signals. The infrasound signals produced by different events have different energy distribution on the frequency axis because of the different mechanism of infrasound excitation. From the characteristics of the monitored infrasound signal, the event type of the infrasound signal can be inferred and the classification and recognition of the infrasound signal can be achieved. The feature extraction and classification recognition of infrasound signal has been a hot topic in the field of infrasound signal processing. It is found that the classification and recognition algorithm of infrasound signal is mainly a breakthrough and development along two main lines. On the one hand, the feature extraction of infrasound signal is studied; on the other hand, the design of pattern recognition algorithm is studied. The key link of infrasound event recognition is the former, that is, how to extract effective features from the signal as the basis for recognition. The recognition effect is essentially determined by the selected pattern features. The research focus of feature extraction is mainly on mining the features that can represent signal categories and how to extract these features effectively. The research focus of this part is on matching the extracted feature vectors, studying the algorithms and structures of various classification models, designing accurate and efficient classifiers, and accomplishing the final goal of accurately classifying the signals. In this paper, we will study the event classification of natural disasters by infrasound signals, focusing on the feature extraction of infrasound signals and the design of pattern recognition classification algorithm. The main contents of this paper are as follows: firstly, the research background and significance of infrasound signal classification and recognition in natural disasters are introduced in detail, and the advantages and disadvantages of various infrasound signal feature extraction algorithms are compared and analyzed. Three technical methods are studied for feature extraction of infrasound signals and two classification algorithms for classification and recognition. Based on this, the complete classification model is verified by using the collected infrasound data. The experimental results are analyzed and the effectiveness of each method is compared. In this paper, the infrasonic signals produced by various events in natural disasters are taken as the research object, with the aim of reducing the false alarm rate in the actual infrasound monitoring, the earthquake, tsunami and volcano are studied. Feature extraction technology and classification recognition algorithm of infrasound signal produced by debris flow. It is expected that the results of this study can be used as a reference for infrasound signal processing methods and practical applications.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN912.3

【参考文献】