TSVD-SCN法提升光纤入侵信号识别性能

54 浏览量更新于2024-08-26 收藏 4.7MB PDF 举报

本文档探讨了一种基于Truncated Singular Value Decomposition (TSVD) 和 Stochastic Configuration Networks (SCN) 的光纤入侵信号识别方法。在光学纤维技术领域，光纤预警告系统对于保障通信安全至关重要。SCN作为一种监督学习模型，通过引入不等式约束来确保模型的泛化能力。然而，SCN在构建过程中存在挑战，主要体现在所使用的激活函数特性以及隐藏节点随机输入权重和偏置的分配方式。这可能导致隐藏输出矩阵变得不稳定，表现为秩不足或多重共线性，从而影响最小二乘法评估输出权重时的性能。传统上，由于这些矩阵问题，SCN在处理数据建模任务时可能表现不佳，其训练和预测能力受限。为了克服这些问题，作者提出了一个创新的方法，即结合TSVD来优化SCN结构。TSVD是一种有效的降维工具，能够减少矩阵的复杂性并提升模型的计算效率，同时有助于解决潜在的线性依赖问题。通过应用TSVD在SCN的隐藏层输出矩阵上，作者旨在提高模型的稳定性和准确性，从而实现更有效的光纤入侵信号识别。具体步骤可能包括以下环节： 1. **矩阵分解**：使用TSVD对隐藏输出矩阵进行分解，提取出最重要的特征向量，这有助于剔除冗余信息，增强模型的表达能力。 2. **参数优化**：在分解后的低秩表示中更新隐藏节点的权重和偏置，利用最小二乘法或其他优化算法找到最佳参数组合，以减少过拟合风险。 3. **模型训练**：使用带有TSVD改进的SCN结构进行训练，确保数据的准确建模，并考虑到监督机制中的不等式约束。 4. **信号识别**：在实际应用中，通过新的SCN模型对光纤传输的信号进行实时分析和分类，以检测潜在的入侵事件。这篇研究论文旨在解决光纤入侵信号识别中的核心问题，通过结合TSVD技术和SCN的特性，提供了一种提高性能和稳定性的新型方法。这种方法不仅在理论上具有理论价值，而且对于实际的光纤安全监控系统具有显著的实际应用潜力。

Contents lists available at ScienceDirect

Optical Fiber Technology

journal homepage: www.elsevier.com/locate/yofte

Optical ﬁber intrusion signal recognition method based on TSVD-SCN

Zhiyong Sheng, Zhiqiang Zeng, Hongquan Qu

⁎

, Yuan Zhang

School of Information Science and Technology, North China University of Technology, Beijing 100144, China

ARTICLE INFO

Keywords:

Stochastic conﬁguration networks

Truncated singular value decomposition

Optical ﬁber pre-warning system

Signal recognition

ABSTRACT

Stochastic Conﬁguration Networks (SCN) introduce the inequality constraint of a supervised mechanism to

ensure the universal approximation property of learner model. However, in the processing of building SCN, due

to the properties of used activation function and the way of assigning the random input weights and biases of the

hidden nodes, the hidden output matrix is often ill-posed, i.e., the matrix can be of rank deﬁcient or demonstrate

multicollinearity. Thus, the least squares method for evaluating the output weights may result in poor gen-

eralization performance for data modelling problems. This paper aims to overcome this drawback through

modifying the computation of the generalized pseudo inverse of the output matrix by a Truncated Singular Value

Decomposition (TSVD) method with an adaptively chosen truncation threshold. The improved SCN model is

then applied for recognizing intrusion signals in Optical Fiber Pre-warning System. Experimental results show

that the proposed improved algorithm can achieve higher recognition rate compared to the original SCN clas-

siﬁer.

1. Introduction

With the rapid development of optical ﬁber sensing technology, the

Optical Fiber Pre-warning System (OFPS) has aroused much attention

in various ﬁelds. The system is characterized by strong resistance to

electromagnetic interference and low construction diﬃculty, and often

applied for inspection of engineering structure safety, perimeter pro-

tection, and pre-warning for oil and gas pipeline safety [1,2]. When the

intrusion behavior appears, the major tasks of the OFPS are to timely

raise alarms and locate the intrusion signal accurately. A lot of re-

searches have been made by researchers on the ﬁber intrusion signal

detection. Some scholars had a further study in the distributed optical

ﬁber vibration sensor [3–5], and document [6] improved the Signal-to-

Noise Ratio (SNR) by using wavelet decomposition de-noising algo-

rithm of optical signals. Qu et al. proposed a Constant False Alarm Rate

(CFAR) algorithm with adaptive detection threshold under diﬀerent

noise backgrounds [7,8], which greatly improved the performance of

OFPS detection algorithms. The above methods adopted the traditional

threshold setting method. However, the problem of false alarm is still

very serious due to the complex work environment of OFPS. Therefore,

it is signiﬁcant to select a new recognition method to identify the ﬁber

signals with the background inference.

Over the past decades, neural network has been widely applied due

to its excellent learning and approximation ability to solve nonlinear

problems [9–12]. However, it is hard to ﬁnd the right tradeoﬀ between

model bias and model variance in building the networks. Practically,

the conventional learning models optimize the network using the error

Back-Propagation (BP) algorithm. Unfortunately, its local minima and

slow convergence will occur when the BP algorithm is performed. In

this case, one of the successful methods is the use of randomized models

[13–16]. Accordingly, the randomized learning techniques for neural

networks aroused the wide attention in the later 80s [17]. In the ﬁeld of

machine learning, stochastic algorithms have great potential to reduce

computational complexity and were further developed in the early 90s

[18,19]. One of the classic models is the Random Vector Functional

Link (RVFL) proposed by [20]. The input weights and biases of RVFL

are selected randomly in accordance with the uniform distribution in a

given range. Besides, the linear regression method is employed to

transform the network which needs multiple iterative optimizations for

a linear least square problem. The above measures greatly reduce the

time required for training the networks. However, the range of random

parameters and the number of hidden layer nodes are directly related to

the performance of RVFL. Tyukin and Prokhorov proved that the ability

of RVFL networks to approximate the objective function is poor if the

above parameters are not properly selected [21]. Wang proposed the

Stochastic Conﬁguration Network (SCN) with randomly selecting the

input weights and biases associated with inequality constraints and

adding the hidden nodes incrementally, which ensuring the universal

approximation property of the models [22–25]. The method of in-

troducing constraints to random algorithms greatly improves the

https://doi.org/10.1016/j.yofte.2019.01.023

Received 14 December 2018; Received in revised form 14 January 2019; Accepted 22 January 2019

⁎

Corresponding author.

E-mail address: qhqphd@ncut.edu.cn (H. Qu).

Optical Fiber Technology 48 (2019) 270–277

Available online 29 January 2019

下载后可阅读完整内容，剩余7页未读，立即下载

weixin_38690545

粉丝: 4
资源: 927

TSVD-SCN法提升光纤入侵信号识别性能

Trust-IPN-TSVD: 动态光散射粒径反演新方法

TTWRIM: 提升动态光散射反演精度的多尺度Tikhonov-TSVD方法

TSVD正则化技术在吉洪诺夫问题中的应用研究

TSVD-Tikhonov正则化多尺度动态光散射反演

theTB-tsvd-675f481.zip_TSVD正则化_refer1cb_tsvd_吉洪诺夫_正则化

论文研究-基于TSVD的图像复制区域伪造检测算法.pdf

基于TSVD求解大地电磁不适定反演问题

基于TSVD的图像复制区域伪造检测算法* (2008年)

svd算法matlab代码-tsvd:基于阈值SVD的K-means用于主题建模

基于张量奇异值分解的人脸识别方法.pdf

最新资源