DMZI振动系统新型端点检测：基于卷积窗高通滤波

145 浏览量更新于2024-08-27 1 收藏 1005KB PDF 举报

"本文提出了一种在DMZI(双马赫-曾德尔)振动系统中基于可配置高通滤波器的端点检测方法，旨在精确识别振动起点。通过利用未干扰噪声的快速傅立叶变换谱，研究人员能够确定合适的截止频率，以此配置高通滤波器来消除主要由低频成分构成的噪声。之后，通过阈值搜索技术对高通滤波器的输出进行分析，从而定位到与振动起始点相对应的时刻。实验结果证明，这种方法在精度、灵活性和可靠性上优于传统的小波检测方法，因此具有广泛的应用前景。" 该文主要探讨了以下几个关键知识点： 1. **端点检测**：端点检测是信号处理中的一个重要概念，目标是识别信号的开始和结束点，对于分析动态系统的行为，如振动系统的特性，尤其重要。本文中，端点检测是为了确定双马赫-曾德尔干涉振动系统的扰动框架的振动起点。 2. **可配置高通滤波器**：高通滤波器允许高频信号通过，而抑制或阻止低频信号，用于去除噪声。文中提到的“可配置”意味着可以根据具体应用和信号特征调整滤波器的截止频率，以优化噪声去除效果。 3. **卷积窗滤波**：这是一种信号处理技术，通过将输入信号与特定的滤波函数（即卷积窗）相乘来改善信号质量。在这里，它用于帮助识别振动的起点。 4. **快速傅立叶变换(FFT)**：FFT是一种高效计算离散傅立叶变换的算法，用于将信号从时域转换到频域，以便分析其频率成分。在文中，FFT用于分析未干扰噪声的频谱，以确定适合的截止频率。 5. **阈值搜索**：这是一种信号处理技术，通过对处理后的信号设置一个阈值，来确定感兴趣的事件或特征。在本文中，阈值搜索用于确定与振动开始点对应的时刻。 6. **小波分析**：作为对比方法，小波分析通常用于检测信号中的局部特征，但实验表明，基于高通滤波器和阈值搜索的方法在端点检测上更有效。这项研究提供了一种新的、高效的振动系统端点检测策略，具有较高的精度和适应性，对于噪声环境下的振动分析具有重要意义，尤其适用于需要精确识别振动起点的工程应用。

1956 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 19, OCTOBER 1, 2014

Conﬁgurable Filter-Based Endpoint Detection

in DMZI Vibration System

Xiangdong Huang, Jia Yu, Kun Liu, Tiegen Liu, and Qinnan Chen

Abstract—In order to accurately determine the vibration

starting point of a disturbance frame in the dual Mach–Zehnder

interferometry vibration system, this letter proposes a novel

endpoint detection method based on convolution window ﬁltering.

Through analyzing the fast Fourier transform spectrum of the

undisturbed noise, we can acquire the cutoff frequency parameter

to conﬁgure one high-pass ﬁlter to remove the noise, which

mainly consists of low-frequency components. Then, the time

instant corresponding to the vibration starting point can be

detected through applying a thresholding search to the output

of the high-pass ﬁlter. Experiments show that the proposed

detection method outperforms the wavelet-based method. Hence,

the proposed detection method has wide application values for

its high precision, ﬂexibility, and reliability.

Index Terms—Endpoint detection, conﬁgurable high-pass

ﬁlter, convolution window, thresholding search.

I. INTRODUCTION

OWADAYS, several distributed optical ﬁber sensors, such

as Dual Mach–Zehnder Interferometry (DMZI) vibration

sensor [1], [2], sagnac interferometry vibration sensor and the

sensor based on optical time-domain reﬂectometer (OTDR)

are widely applied in disturbance detection. Among these,

DMZI vibration sensor receives increasing attention, for it

has the superiority of high sensitivity and fast response. Thus

it is widely applied in external invasion event detection,

such as pipeline leakage detection [3], perimeter security [4],

submarine cable security [5] etc. Hence, determination of

these events’ sudden moments is of vital importance in

DMZI vibration system. As far as the collected vibration

samples are concerned, vibration starting moment actually

corresponds to some endpoint, i.e., the boundary between the

undisturbed noise and the disturbance signal. Furthermore,

endpoint detection is also the basis of other techniques in

DMZI system, such as event positioning [6], [7] and event dis-

crimination [8], which are all realized through processing the

samples following the detected endpoint. As [9] demonstrates,

endpoint detection also contributes to enhancing the weight

Manuscript received June 8, 2014; revised July 11, 2014; accepted

July 21, 2014. Date of publication July 29, 2014; date of current version

September 8, 2014. This work was supported in part by the National Basic

Research Program of China under Grant 2010CB327806 and in part by the

National Natural Science Foundation of China under Grant 61271069 and

Grant 61271322. (Corresponding author: Kun Liu.)

X. Huang and J. Yu are with the School of Electronic Informa-

tion Engineering, Tianjin University, Tianjin 300072, China (e-mail:

xdhuang@tju.edu.cn; jia7721927@126.com).

K. Liu, T. Liu, and Q. Chen are with the College of Precision Instrument

and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China

(e-mail: beiyangkl@tju.edu.cn; tgliu@tju.edu.cn; chenqinnan@gmail.com).

Color versions of one or more of the ﬁgures in this letter are available

online at http://ieeexplore.ieee.org.

Digital Object Identiﬁer 10.1109/LPT.2014.2343274

of high-frequency components as well as attenuating the

magnitude of low-frequency main lobe, which contributes to

reduce position error. Thus, the accuracy of endpoint detection

will inevitably inﬂuence the position error and recognition rate

of events (such as climbing the fence, cable invasion).

In recent years, various endpoint detection algorithms,

such as cepstrum coefﬁcients method [10], short-time zero

crossing ratio method [11], wavelet-based method [12], [13],

have been proposed [14]. However, cepstrum coefﬁcients

method only applies to the occasion when multiplicative

noise instead of additive noise is involved, so it is only

applied in some special ﬁelds such as speech and radar

systems. As to the short-time zero crossing ratio method,

which is based on counting the numbers of zero crossing in

some ﬁxed time blocks, it actually determines the position

where the highest frequency component occurs. In fact, there

exists some distance between this position and the endpoint.

It should be emphasized, since the wavelet can be easily

zoomed out by simply choosing some ﬁne scale so that the

high-frequency detail information of the sudden event can

be extracted, the wavelet-based method is the mainstream

in endpoint detection. However, wavelet based method also

cannot distinguish undisturbed noise and disturbance signal

accurately, since wavelet decomposition can only split the

frequency band in half. Thus wavelet decomposition has to

be implemented in high levels to ensure that the observed

frequency band can approximate the boundary frequency

between the undisturbed noise and disturbance signal as

close as possible, which also introduces high computation

complexity and makes it difﬁcult to detect the endpoint in

real-time.

In fact, since the core problem of endpoint detection

is to remove undisturbed noise, whose energy is mainly

distributed in low frequency band, it is essential to accu-

rately analyze the frequency band of the undisturbed noise.

So this letter puts forward one novel detection algorithm

based on high-pass convolution window FIR ﬁlter, whose

cutoff frequency is acquired from the FFT result of undis-

turbed noise. This letter will also interpret the principle of

convolution window based ﬁlter, whose parameters can be

conﬁgured ﬂexibly to apply to different application envi-

ronments. In addition, comparing with the existing high-

pass digital ﬁlters such as Butterworth and Chebyshev ﬁlters,

the proposed ﬁlter is permanently stable. Experiments also

show that the proposed endpoint detection method outper-

forms the wavelet based method in accuracy, ﬂexibility, and

reliability.

II. DMZI S

YSTEM

A DMZI vibration sensor is shown in Fig. 1: the output of

the laser with narrow linewidth is split equally at coupler C1

See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

下载后可阅读完整内容，剩余3页未读，立即下载

weixin_38670433

粉丝: 9
资源: 899

DMZI振动系统新型端点检测：基于卷积窗高通滤波

基于FPGA的语音端点检测.pdf

基于可配置滤波器的DMZI振动系统端点检测

基于熵函数的语音端点检测方法.rar_c 端点检测_基于熵 语音 端点 检测_端点检测_语音 检测_语音 端点检测

edp.rar_EPD 端点检测_edp_edp版本检测_基于 MATLAB 语音信号 端点检测_端点检测

基于谱熵语音端点检测

基于FPGA的语音端点检测

基于倒谱特征的端点检测

matlab 基于双门限的端点检测

基于端点检测和高斯滤波器组的MFCC说话人识别1

基于MATLAB的语音端点检测.doc

最新资源

基于熵函数的语音端点检测方法.rar_c 端点检测_基于熵语音端点检测_端点检测_语音检测_语音端点检测

edp.rar_EPD 端点检测_edp_edp版本检测_基于 MATLAB 语音信号端点检测_端点检测