A method for recognizing vibration signals has been developed

A method for recognizing vibration signals has been developed

In an article published in Sensorsresearchers proposed a new endpoint detection algorithm and a feature fusion technique based on two-dimensional convolutional neural network (2DCNN) based on multidimensional feature fusion to identify fiber vibration signals.

Studies: Recognition of optical fiber vibration signals based on multi-stage feature fusion. Image credit: AerialVision_it/Shutterstock.com

The Sagnac-type distributed fiber optic sensing system suffered from the disadvantage of low accuracy in detecting disturbance vibration events. The method demonstrated in the current paper combines conventional fiber optic vibration signal recognition with automatic feature fusion using 2DCNN to achieve higher accuracy.

A new endpoint detection algorithm was used to detect the vibrational component of the original signal to increase the endpoint detection efficiency. This algorithm combined the energy spectral entropy and the product of the spectral centroid to achieve the desired detection effect.

Multiscale and multiscale signal features were extracted using 2DCNN of different scales. A new technique for integrating differential pooling functions was used to deal with information loss during the pooling process.

The extracted features were recognized using a multilayer perceptron (MLP). Experiments showed that the efficiency of distributed optical fiber detection systems was higher compared to conventional variational mode decomposition (VMD) and empirical mode decomposition (EMD) and one-dimensional CNN (1D-CNN) approaches.

The growing need for reliable perimeter security systems

The requirements for safe and reliable perimeter security systems are becoming more and more urgent due to the rapid technological and scientific progress and the growth of living standards. Thus, fiber optic perimeter security has attracted considerable interest from researchers around the world, a technology with significant growth potential.

Fiber optic perimeter security systems have been widely used in tunnel detection, border monitoring, offshore oil exploration, seismic monitoring, marine monitoring, perimeter security and pipeline monitoring.

The distributed optical fiber sensing system based on the Sagnac interferometer has small dimensions, high sensitivity and easy installation.

Due to the basic construction and symmetrical design of the distributed optical fiber sensing system, fiber vibration signal detection is suitable for distributed deployment without the need for a high degree of light source coherence. There is no requirement for a reference thread. Therefore, in the current research, the recognition of a Sagnac-type optical fiber distributed sensor was carried out.

The three basic processes in the identification of optical fiber vibration signals are feature extraction, classification, and preprocessing of optical fiber vibration signals. The primary goal of preprocessing is vibration signal noise suppression and endpoint detection.

Conventional fiber optic vibration signal recognition is primarily limited to its own method of feature extraction and subsequent identification and classification, which frustrates the relationship between the two and leads to the loss of certain signal information. A single-stage CNN can only extract part of the signal information and is unable to extract data from different signal scales.

In this paper, the researchers demonstrated a distributed fiber optic sensing technique for identifying fiber optic vibration signals. The endpoint detection effect was effectively improved by a new endpoint detection algorithm that combined energy spectral entropy and spectral centroid products. This algorithm also extracted features at different scales and levels of 2DCNN. It used a new technique for differential feature pooling and feature fusion to further increase accuracy in distributed fiber optic sensing.

Construction of an improved vibration signal recognition mechanism

Signal preprocessing, multilevel feature fusion extraction, differential pooling function, and classification recognition formed the basic framework of optical fiber vibration signal detection based on multilevel feature fusion.

Many ineffective silent signals were found in unspecified optical fiber vibration signal data. Therefore, detection of the endpoints of the original optical fiber vibration signals was performed before creating the experimental data set. The vibration part of the signal was then captured to reduce the analysis of silent signals and increase the operating speed. The detection of the end point of the signal was primarily used to detect the start and end points of the vibration part of the signal.

2DCNN classified and recognized the pre-processed optical fiber vibration signal to more comprehensively represent the exhaustive information of the original signal, increase the recognition speed and minimize the error rate.

2DCNN required less parameter consideration than other shallow or deep neural networks, and the hidden layers of 2DCNN processed the input data in a filtered form. The 2DCNN network could intelligently extract the relevant features from the convolutional and differential pooling function layers and automatically reduce the noise in the signal.

This work proposed a differential pooling function structure that integrated average pooling and maximum pooling to make the features of the recovered vibration signal more thorough and discriminative, which increases the recognition accuracy of the optical fiber vibration signal. Differential pooling features minimized the impact of background noise on the signal while collecting specific information about the fiber vibration signal, thus combining the benefits of peak pooling and average pooling.

Comparative experiments were conducted using the collected datasets to confirm the effectiveness of the fiber optic distributed sensing method. In the experiments of this work, the multi-level feature fusion technique (Fusion 2DCNN) was compared with the first channel CNN (First 2DCNN), the two-channel CNN (FS 2DCNN) and the second channel CNN (Second 2DCNN). .

The performance of the multi-scale feature fusion (Fusion 2DCNN) approach was better than that of the FS 2DCNN method, demonstrating that the differential feature fusion approach combines the advantages of average pooling and maximum pooling. Thus, it was observed that the differential feature pooling approach was critical for enhancing the performance of distributed fiber optic sensing.

New recognition of vibration signals and a safer tomorrow

In this paper, an innovative fiber vibration signal identification method based on a novel endpoint detection technique and a multi-scale feature fusion approach was proposed. The endpoint detection algorithm was based on the combination of energy spectral entropy and spectral centroid product and merging their respective advantages. Therefore, the endpoint detection method effectively improved the detection accuracy of optical fiber vibration signals and had an improved detection impact on low-frequency signals.

The running and walking signals, which were difficult to distinguish due to the short vibration time, could be effectively distinguished by the signal captured for 1 s. A multi-scale feature fusion method was used to more effectively extract multi-scale and multi-scale information about the features of fiber optic vibration signals and to classify and recognition were used by MLP.

A comparison of conventional VMD and EMD pattern recognition techniques, 1D-CNN and multi-level fusion methods showed that the average accuracy of the method proposed in this work was higher than other traditional methods. Therefore, the distributed optical fiber sensing system can effectively reduce the false alarms of optical fiber vibration signal recognition and accurately identify the four kinds of optical fiber vibration signals of knocking, running, walking and waving.


X. Ma, J. Mo, J. Zhang, J. Huang, Optical fiber vibration signal recognition based on multi-scale feature fusion. 2022. Sensors.http://www.mdpi.com/1424-8220/22/16/6012/htm

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork, the owner and operator of this website. This disclaimer forms part of the terms of use of this website.

#method #recognizing #vibration #signals #developed

Leave a Comment

Your email address will not be published. Required fields are marked *