Classification Of Acoustic Signals Research Articles

Personalized federated learning on NLOS acoustic signal classification

AbstractTo identify non‐line‐of‐sight, acoustics‐based indoor positioning needs to collect audio recordings of sound fields in multiple rooms and upload them to the central server for training. However, if the transmission process or the server‐side suffers from malicious attacks, private data may be leaked. To address both training difficulties and privacy issues simultaneously, a novel model for personalized federated learning is proposed. This model incorporates user frequency and room data capacity while also taking into account significant differences in positioning data and room layout. The proposed model can accurately identify differences between various room data during aggregation on the server‐side. By collecting data in actual indoor environments and comparing it with existing algorithms, the proposed method achieved 90% accuracy in verifying data for unfamiliar rooms.

PLDA in i-vector based underwater acoustic signals classification

ABSTRACT Intelligent recognition of underwater acoustic targets is crucial for exploiting marine resources. In this paper, we propose a deep learning method for underwater acoustic signal recognition consisting of three steps: Firstly, we extract the underwater acoustic signal's identification vector (i-vector) using a Gaussian mixture model. Secondly, we construct a probabilistic linear discriminant analysis (PLDA) model that divides the i-vector into two segments, Finally, we used a multi-long short-term memory neural network (MLSTMNN) to learn the features, outputting the underwater acoustic signal labels through the softmax layer. Experiments using the ShipsEar dataset can obtain a recognition rate of 84.8%; compared to the traditional Mel frequency Cepstral Coefficients (MFCC) and convolutional neural networks, the recognition rate greatly improved.

Feature extraction, recognition, and classification of acoustic emission waveform signal of coal rock sample under uniaxial compression

In this study, based on Mel frequency cepstrum coefficient (MFCC) method, the AE signal characteristics of coal and rock samples were extracted, and the stress state criterion based on signal features was constructed. By integrating back propagation (BP) neural network for deep learning of signal characteristics, the recognition, classification, and prediction of coal and rock materials were realized. The results show that the MFCC could characterize the variation law of the original signal, with the sharp fluctuation of the amplitudes of both the AE signal and MFCC when the rock stress was near the peak value. Considering the ratio of sample stress to peak stress as the stress state, the correlation between MFCC and stress state was analyzed. The BP neural network exhibited a high accuracy rate for the signal characteristics represented by MFCC, achieving an accuracy of more than 95% with a fast recognition speed. Notably, the evaluation results of neural network model were stable and reliable. Therefore, MFCC can be used to extract the AE waveform signal characteristics and evaluate the stability of stress state for coal and rock materials. The recognition, classification, and prediction of high-precision results of the two types of waveform characteristics of coal and rock can be achieved through BP neural network.

Identifying the Acoustic Source via MFF-ResNet with Low Sample Complexity

Acoustic signal classification plays a central role in acoustic source identification. In practical applications, however, varieties of training data are typically inadequate, which leads to a low sample complexity. Applying classical deep learning methods to identify acoustic signals involves a large number of parameters in the classification model, which calls for great sample complexity. Therefore, low sample complexity modeling is one of the most important issues related to the performance of the acoustic signal classification. In this study, the authors propose a novel data fusion model named MFF-ResNet, in which manual design features and deep representation of log-Mel spectrogram features are fused with bi-level attention. The proposed approach involves an amount of prior human knowledge as implicit regularization, thus leading to an interpretable and low sample complexity model of the acoustic signal classification. The experimental results suggested that MFF-ResNet is capable of accurate acoustic signal classification with fewer training samples.

Deep learning-based classification of damage-induced acoustic emission signals in UHPC

Acoustic emission (AE) is advantageous in studying and monitoring the damage state of UHPC, and its performance can be further improved if various damage-induced AE signals are classified accurately. The deep learning method is promising to enhance the classification accuracy, but it is limited by the absence of the corresponding AE dataset. In this study, fiber pullout tests and direct tension tests (DTTs) were performed to explore the time–frequency features of various damage-induced AE signals in UHPC. The AE sources can be classified into five categories: matrix cracking, fiber debonding, fiber sliding, fiber scraping, and matrix spalling. A corresponding AE dataset was established after manual pre-labeling, pseudo-labeling with transfer learning, and mix-up data augmenting. Several lightweight convolutional neural networks (CNNs) were trained from scratch based on the dataset, with modified time–frequency spectrograms as data input. The performances of CNN-based classification models are superior to traditional shallow-learning methods. The popular lightweight CNN structure of ResNet18 achieves the highest overall accuracy of 93.94% among adopted CNN structures such as GoogleNet, ResNet18, EfficientNet-b0, and MobileNetV2. The AE signal classification results of DTTs show that fiber-pullout damage and matrix cracking occur intensively at the strain-hardening phase and increase with fiber content. That suggests the increasing fiber bridging and multi-crack propagation in the uniaxial tension of UHPC with higher fiber content.

Novel DEMON Spectra Analysis Techniques and Empirical Knowledge Based Reference Criterion for Acoustic Signal Classification

This paper presents some novel methods to estimate a vessel’s number of shafts, course, speed and classify it using the underwater acoustic noise it generates. A classification framework as well as a set of reference parameters for comparison are put forth. Identifying marine traffic in surroundings is an important task for vessels in an open sea. Vessels in vicinity can be identified using their signatures. One of the typical signatures emitted by a vessel is its acoustic measurements. The raw sonar data consisting of the acoustic signatures is generally observed manually by sonar operators for suggesting class of query vessel. The valuable information that can be extracted from the recorded acoustic signature includes shaft revolutions per minute (SRPM), number of blades (NOB), number of shafts, course and speed etc. Expert sonar operators use their empirical knowledge to estimate a vessel’s SRPM and NOB. Based on this information vessel classification is performed. Empirical knowledge comes with experience, and the manual process is prone to human error. To make the process systematic, calculation of the parameters of the received acoustic samples can be visually analyzed using Detection of Envelope Modulation on Noise (DEMON) spectra. Reported research mostly focuses on SRPM and NOB. Parameters such as number of shafts and vessel course and speed can effectively aid the vessel classification process. This paper makes three novel contributions in this area. Firstly, some novel DEMON spectra analysis techniques are proposed to estimate a water vessel’s number of shafts, speed, and relative course. Secondly, this paper presents a classification framework that uses the features extracted from DEMON spectra and compares them with a reference set. Thirdly, a novel set of reference parameters are provided that aid classification into categories of large merchant ship type 1, large merchant ship type 2, large merchant ship type 3, medium merchant ship, oiler, car carrier, cruise ship, fishing boat and fishing trawler. The proposed analysis and classification techniques were assessed through trials with 877 real acoustic signatures recorded under varying conditions of ship’s speed and sea state. The classification trials revealed a high accuracy of 94.7%.

Recognition of behavior state of Penaeus vannamei based on passive acoustic technology

Underwater signal processing based on passive acoustic technology has carried out a lot of research on the behavioral sound of underwater creatures and the protection of marine resources, which proves the feasibility of passive acoustic technology for long-term and non-destructive monitoring of underwater biological sound production. However, at present, most relevant research focuses on fish but little research on shrimp. At the same time, as the main economic breeding industry, Penaeus vannamei has a backward industrial structure, in which the level of intelligence needs to be improved. In this paper, the acoustic signals generated by different physiological behaviors of P. vannamei are collected based on passive acoustic technology. Their different behaviors are finally classified and identified through feature extraction and analysis. Meanwhile, the characteristic non-parametric ANOVA is carried out to explore the relationship between the acoustic signals and the behavior state of P. vannamei to achieve the purpose of real-time monitoring of the behavior state of P. vannamei. The experimental results show that linear prediction cepstrum coefficient (LPCC) and Mel-frequency cepstrum coefficient (MFCC) characteristic coefficients are effective in the classification and recognition of different behavioral acoustic signals with interspecific acoustic signals of P. vannamei. Meanwhile, the SVM classifier based on OvR classification strategy can model the acoustic signal characteristics of different underwater biological behaviors more efficiently and has classification accuracy as high as 93%.

Acoustic Signal Classification by Support Vector Machine for Pipe Crack Early Warning in Smart Water Networks

Uncontrolled pipe breaks are a challenge for water utilities all over the world. This paper describes a technique that enables pipe cracks to be identified at an early stage before they become uncontrolled breaks by utilizing a permanent acoustic monitoring system as part of a smart water network. Multiple acoustic features are selected and extracted from recorded wave files that are associated with proactive repair and uncontrolled pipe break events. The extracted acoustic features and the associated wave file labels (as either crack/leak noise or no crack/leak noise) are used to train a support vector machine model. The trained model has been operationalized in the South Australia Water Corporation’s smart water network analytics platform to process incoming new acoustic wave files in a near-real-time manner. If the acoustic wave file is classified as a pipe crack/leak, an alarm is sent to an investigation crew such that leak localization can be conducted and repairs started. The successful detection of multiple pipe cracks/leaks by the developed model after its implementation proves that it is an effective tool to enable proactive management of pipe breaks in water distribution systems.

Improved Methodology of SVM to Classify Acoustic Signal by Spectral Centroid

Acoustic signal classification issues are addressed in this work using spectral examination, channel extracting the features from the input and machine learning algorithm. This brief article examines the effect of various settings on feature extraction. This feature-level channel combination's accuracy increase is then observed. To categorise things, pattern recognition utilises a variety of classification schemes. "Pattern" refers to the measures that must be categorised with accurate feature extracted. Images and audio signals are among the most common kinds of measurements. The proposed Support Vector Machine (SVM) is used for the necessity of an effective categorization of acoustic signals driven by the continual improvements in multimedia technology. This study uses two machine learning algorithms to enhance audio classification and categorization. The proposed SVM achieves superior performance than the other ML algorithm by spectral features.

Acoustic Signal Classification Using Symmetrized Dot Pattern and Convolutional Neural Network

The classification of sound signals can be applied to the fault diagnosis of mechanical systems, such as vehicles. The traditional sound classification technology mainly uses the time-frequency domain characteristics of signals as the basis for identification. This study proposes a technique for visualizing sound signals, and uses artificial neural networks as the basis for signal classification. This feature extraction method mainly uses a principle to convert a time domain signal into a coordinate symmetrized dot pattern, and presents it in the form of snowflakes through signal conversion. To verify the feasibility of this method to classify different noise characteristic signals, the experimental work is divided into two parts, which are the identification of traditional engine vehicle noise and electric motor noise. In sound measurement, we first use the microphone and data acquisition system to measure the noise of different vehicles under the same operating conditions or the operating noise of different electric motors. We then convert the signal in the time domain into a symmetrized dot pattern and establish an acoustic symmetrized dot pattern database, and use a convolutional neural network to identify vehicle types. To achieve a better identification effect, in the process of data analysis, the effect of the time delay coefficient and weighting coefficient on the image identification effect is discussed. The experimental results show that the method can be effectively applied to the identification of traditional engine and electric vehicle classification, and can effectively achieve the purpose of sound signal classification.

A Speech recognition of isolated words in the “Baoulé” language via back-propagation neural networks (BNN)

The main objective of this research is to explore how a back propagation neural network (BNN) can be applied to the speech recognition of isolated words. The simulation results show that a BNN offers an efficient approach for small vocabulary systems. The recognition rate reaches 100% for a 5 word system and 94% for a 10 word system. The general techniques developed in this article can be extended to other applications, such as the recognition of sonar targets and the classification of underwater acoustic signals.

Earthquake source characterization by machine learning algorithms applied to acoustic signals

Underwater seismic events generate acoustic radiation (such as acoustic-gravity waves), that carries information about the source and can travel long distances before dissipating. Effective early warning, emergency response, and information dissemination for earthquakes and tsunamis require a rapid characterisation of the fault properties: geometry and dynamics. In this work, we analysed hydrophone recordings of 201 earthquakes, located in the Pacific and the Indian Ocean, by employing acoustic signal processing and classification methods. The analysis allows identifying the type of earthquake (i.e. slip type, magnitude) and provides near real-time estimation of the effective properties of the fault dynamics and geometry. The results were compared against values reported by the Harvard Global Centroid Moment Tensor catalog (gCMT), revealing statistical significance between the extracted acoustic properties used to feed machine learning algorithms and the predicted slip and magnitude values.

Classification of acoustic emission signals in wood damage and fracture process based on empirical mode decomposition, discrete wavelet transform methods, and selected features

The nondestructive testing technology of generated acoustic emission (AE) signals for wood is of great significance for the evaluation of internal damages of wood. To achieve more accurate and adaptive evaluation, an AE signals classification method combining the empirical mode decomposition (EMD), discrete wavelet transform (DWT), and linear discriminant analysis (LDA) classifier is proposed. Five features (entropy, crest factor, pulse factor, margin factor, waveform factor) are selected for classification because they are more sensitive to the uncertainty, complexity, and non-linearity of AE signals generated during wood fracture. The three-point bending load damage experiment was implemented on sample wood of beech and Pinus sylvestris to generate original AE signals. Evaluation indexes (precision, accuracy, recall, F1-score) were adopted to assess the classification model. The results show that the ensemble classification accuracies of two tree species reach 94.58% and 90.58%, respectively. Moreover, compared with the results of the original AE signal, the accuracy of the AE signal processed by the methods proposed is increased by 27.68%. It indicates that the EMD and DWT signal processing methods and selected features improve the classification accuracy, and this automatic classification model has good AE signal recognition performance.

Wood acoustic emission signals classification based on pseudospectrum, and entropy

The nondestructive testing technology of generated acoustic emission(AE) signals for wood is of great significance for the evaluation of internal damages of wood. In order to improve the classification accuracy and adaptability of AE signal, we selected two features(pseudospectrum, entropy) for classify AE signals in the process of wood fracture using SVM classifier. The three-point bending load damage experiment was utilized to generate original AE signals. Evaluation indexes(Precision, Accuracy, Recall, F1-score, Cohen Kappa score, Matthews Corrcoef) were adopted to assess the classification model. The results showed that the overall accuracy of the SVM classification model obtained by the method combining pseudospectrum and entropy features is 89.44%, which indicates that this automatic classification model has good AE signal recognition performance.

Investigating and improving the utility of probabilistic linear discriminant analysis for acoustic signal classification

Probabilistic linear discriminant analysis (PLDA) has achieved good performance in face recognition and speaker recognition. However, the computation of PLDA using the original formulation is inefficient when there are many training data, especially when the dimensionality of the data is high. Faced with this inefficiency issue, we propose scalable formulations for PLDA. The computation of PLDA using the scalable formulations is more efficient than using the original formulation when dealing with many training data. Using the scalable formulations, the PLDA model can significantly outperform other popular classifiers for speaker recognition, such as Support Vector Machine (SVM) and Gaussian Mixture Model (GMM). Besides of directly using PLDA as a classifier, we may also use PLDA as a feature transformation technique. This PLDA-based feature transformation technique can reduce or expand the original feature dimensionality, and at the same time keep the transformed feature vector approximately following the Gaussian distribution. Our experimental results on speaker recognition and acoustic scene classification demonstrate the effectiveness of applying PLDA for feature transformation. It is then promising to combine PLDA with other classification models for improved performance, extending the utility of PLDA to a wider range of areas.

Sound classification during Seabed Classification Experiment 2017

In ocean acoustics, finding acoustic signals within long recordings is usually time consuming. The need to optimize this process we propose an experiment to discover the optimal acoustic signal classification model using the PyTorch deep learning package. This machine learning algorithm was designed to recognize and classify various sources from an input of spectrograms compiled from raw sound data. In the case of continuous audio files, this model’s purpose is to identify areas of interest for human operators so they don’t need to listen to each hour of the audio files marking down noises as they are heard. Four Convolutional Neural Networks (CNN) with differing numbers of layers took in one minute spectrograms. The most successful of the CNNs acted on time-averaged spectrograms, and resulted in a model which achieved a high degree of accuracy. This machine learning algorithm can help identify underwater sound signal sources, and can more efficiently identify when different signals are present in long audio files. The results of these tests imply that time averaging spectrograms may improve the identification of long term signal sources by a CNN. [ Research supported by the NSF REU program.]

A method for underwater acoustic signal classification using convolutional neural network combined with discrete wavelet transform

The detection and classification of underwater targets such as fish are one of the major tasks of the underwater acoustic signal processing and are very important for scientific, fisheries and ocean engineering and economic fields. The convolutional neural network (CNN) combined with the discrete wavelet transform (DWT) (namely CNN_DWT) not only reduces the data processing dimension of signals and the computational costs of the signal analysis, but also improves the performance of target detection and classification. This paper proposes a new CNN to classify the images that reflected the underwater acoustic signal in the database that is made up of the scalogram of underwater acoustic signals. Also, in order to attain greater accuracy and comparable efficiency to the spatial domain processing, we convert the data to the wavelet domain. Also, we propose a deep learning method for the classification of underwater acoustic signals using the new CNN combined with DWT. Next, through the simulation experiment, we evaluate our new method for underwater acoustic signal classification using the CNN combined with DWT, by comparing with classical methods. Comparing the proposed method to spatial domain CNN and classical methods, the experimental results reveal a substantial increment in classification accuracy and noise robustness. And the learning curves show that the proposed CNN_DWT does not generate the overfitting problem and its generalization ability is high. The proposed CNN_DWT improves the classification accuracy and convergence of underwater acoustic signals than the classical CNNs. The noise robustness of the proposed CNN_DWT is higher than those of classical CNNs and back-propagation neural networks (BPNNs) for the classification of underwater acoustic signals. Experimental results show that the classification performance of new CNN combined with DWT is higher than those of classical CNNs and BPNNs for the classification of underwater acoustic signals.

Underwater Acoustic Signal Classification Based on Sparse Time–Frequency Representation and Deep Learning

For classification of underwater acoustic signals, we propose a novel sparse anisotropic chirplet transform (ACT) to reveal fine time-frequency structures. The signal features in the form of a time-frequency map are fed into a deep convolutional neural network, referred to as a time-frequency feature network (TFFNet), which brings flexibility to signal classification. The TFFNet is based on a novel efficient feature pyramid enhancing feature (EFP) maps by aggregating the context information at different scales. To remove the gridding artifacts on enhanced feature maps, a form of aggregating transformation, a forward feature fusion, is utilized to merge the forward feature maps. Main contributions of this work are a novel sparse ACT, a TFFNet classifier, and an EFP with forward feature fusion. Experimental results demonstrate that the sparse ACT provides a high-resolution time-frequency representation of underwater signals and the TFFNet improves the classification performance compared to known networks and two machine learning methods (random forest and support vector machine with radial basis function kernel) on two real data sets, an underwater acoustic communication signal data set and whale sounds data set.

Acoustic Emission Signal Classification Using Feature Analysis and Deep Learning Neural Network

With the development of artificial intelligence in recent years, deep learning has been widely used in mechanical system signal classification but the impact of different feature extractions on the efficiency and effectiveness of deep learning neural networks is more important. In this study, a vehicle classification based on engine acoustic emission signal in the time domain, the frequency domain and the wavelet transform domain for deep learning network techniques is presented and compared. In signal classification, different feature extractions will show in different decomposition levels and can be used to recognize the various acoustic conditions. In the experimental work, as engines from 10 different ground vehicles operate, the measured sound signal is converted into a digital signal, and the established data set is classified and identified by the deep learning method. The number of samples, identification rate and identification time in the various signal domains are compared and discussed in this study. Finally, the experimental results and data analysis show that by using the wavelet signal and the deep learning method, excellent identification time and identification rate can be achieved, compared with traditional time and frequency domain signals.

Method of Analysis and Classification of Acoustic Emission Signals to Identify Pre-Seismic Anomalies

Method of Analysis and Classification of Acoustic Emission Signals to Identify Pre-Seismic Anomalies