Search

Purpose: This paper aims to address the challenges in hydroacoustic signal detection, signal distortion and target localization caused by baseline drift. The authors propose a combined algorithm that integrates short-time Fourier transform (STFT) detection, smoothness priors approach (SPA), attitude calibration and direction of arrival (DOA) estimation for micro-electro-mechanical system vector hydrophones. Design/methodology/approach: Initially, STFT method screens target signals with baseline drift in low signal-to-noise ratio environments, facilitating easier subsequent processing. Next, SPA is applied to the screened target signal, effectively removing the baseline drift, and combined with filtering to improve the signal-to-noise ratio. Then, vector channel amplitudes are corrected using attitude correction with 2D compass data. Finally, the absolute target azimuth is estimated using the minimum variance distortion-free response beamformer. Findings: Simulation and experimental results demonstrate that the SPA outperforms high-pass filtering in removing baseline drift and is comparable to the effectiveness of variational mode decomposition, with significantly shorter processing times, making it more suitable for real-time applications. The detection performance of the STFT method is superior to instantaneous correlation detection and sample entropy methods. The final DOA estimation achieves an accuracy within 2°, enabling precise target azimuth estimation. Originality/value: To the best of the authors' knowledge, this study is the first to apply SPA to baseline drift removal in hydroacoustic signals, significantly enhancing the efficiency and accuracy of signal processing. It demonstrates the method's outstanding performance in the field of underwater signal processing. In addition, it confirms the reliability and feasibility of STFT for signal detection in the presence of baseline drift. [ABSTRACT FROM AUTHOR]

This study introduces the Synchro-Reassigning Transform (SRT), a cutting-edge method developed for high-resolution time-frequency analysis of power quality (PQ) disturbances. The research assesses the performance of SRT in comparison to other established techniques such as Short-Time Fourier Transform (STFT), Wavelet Transform (WT), Synchro-Squeezing Transform (SST), and Synchro-Extracting Transform (SET). Specifically, the study focuses on the identification and categorization of various PQ issues, including voltage sags, swells, interruptions, harmonics, and inter-harmonics. The analysis reveals that SRT excels in detecting harmonics and inter-harmonics, providing much clearer and more detailed time-frequency representations by minimizing cross-term interference, which is a limitation often observed in other methods. The results highlight SRT's ability to offer more accurate and reliable signal interpretations, leading to enhanced precision in PQ analysis. Overall, this technique represents a significant advancement in power quality monitoring, delivering greater reliability and improving the accuracy of PQ disturbance detection. [ABSTRACT FROM AUTHOR]

Epilepsy, a neurological disease characterized by recurrent seizures, can be diagnosed using Electroencephalogram (EEG) signals. Traditional diagnostic methods often face limitations, leading to delays and potential misdiagnoses. In response, researchers have been developing low-cost assistive systems to enhance diagnostic accuracy and reduce life-threatening risks for epilepsy patients. In this study, a hybrid approach is proposed to diagnose epilepsy disease. To validate the success of the proposed algorithm, Hauz Khas and Bonn data sets were used. AlexNet, GoogleNet, VGG19, ResNet50, and ResNet101 classifiers were employed in this study along with the Continuous Wavelet Transform (CWT) and Short Time Fourier Transform (STFT). To increase the generalization capability, 10-fold cross-validation method was used in the classification process. Firstly, the preictal and ictal moments in the Hauz Khas dataset was classified with 99.5% success rate by CWT method and Resnet101. Similarly, 99.8% accuracy was achieved in the binary classification of the Bonn dataset using the CWT method with Resnet101. Finally, for the classification with the AB-CD-E group, 99.33% classification success rate was achieved by using the CWT method with the Resnet-101 model. These findings underscore the potential of the proposed assistive system to significantly improve the diagnosis and management of epilepsy, demonstrating high accuracy and reliability across different datasets and classification techniques.

Abstract The microarray gene expression data poses a tremendous challenge due to their curse of dimensionality problem. The sheer volume of features far surpasses available samples, leading to overfitting and reduced classification accuracy. Thus the dimensionality of microarray gene expression data must be reduced with efficient feature extraction methods to reduce the volume of data and extract meaningful information to enhance the classification accuracy and interpretability. In this research, we discover the uniqueness of applying STFT (Short Term Fourier Transform), LASSO (Least Absolute Shrinkage and Selection Operator), and EHO (Elephant Herding Optimisation) for extracting significant features from lung cancer and reducing the dimensionality of the microarray gene expression database. The classification of lung cancer is performed using the following classifiers: Gaussian Mixture Model (GMM), Particle Swarm Optimization (PSO) with GMM, Detrended Fluctuation Analysis (DFA), Naive Bayes classifier (NBC), Firefly with GMM, Support Vector Machine with Radial Basis Kernel (SVM-RBF) and Flower Pollination Optimization (FPO) with GMM. The EHO feature extraction with the FPO-GMM classifier attained the highest accuracy in the range of 96.77, with an F1 score of 97.5, MCC of 0.92 and Kappa of 0.92. The reported results underline the significance of utilizing STFT, LASSO, and EHO for feature extraction in reducing the dimensionality of microarray gene expression data. These methodologies also help in improved and early diagnosis of lung cancer with enhanced classification accuracy and interpretability.

AbstractAccurate detection of exercise fatigue based on physiological signals is vital for reasonable physical activity. Existing studies utilize widely Electrocardiogram (ECG) signals to achieve exercise monitoring. Nevertheless, ECG signals may be corrupted because of sweat or loose connection. As a non-invasive technique, Phonocardiogram (PCG) signals have a strong ability to reflect the Cardiovascular information, which is closely related to physical state. Therefore, a novel PCG-based detection method is proposed, where the feature fusion of deep learning features and linear features is the key technology of improving fatigue detection performance. Specifically, Short-Time Fourier Transform (STFT) is employed to convert 1D PCG signals into 2D images, and images are fed into the pre-trained convolutional neural network (VGG-16) for learning. Then, the fusion features are constructed by concatenating the VGG-16 output features and PCG linear features. Finally, the concatenated features are sent to Support Vector Machines (SVM) and Linear Discriminant Analysis (LDA) to distinguish six levels of exercise fatigue. The experimental results of two datasets show that the best performance of the proposed method achieves 91.47% and 99.00% accuracy, 91.49% and 99.09% F1-score, 90.99% and 99.07% sensitivity, which has comparable performance to an ECG-based system which is as gold standard (94.32% accuracy, 94.33% F1-score, 94.52% sensitivity). [ABSTRACT FROM AUTHOR]

Computerized evaluation of valve anomalies from cardiac sound is a well-tried endeavor in cardiology. Conversely, automated methods for the diagnosis of cardiovascular diseases mainly depend on the features collected from the cardiac signal. Analyzing phonocardiogram (PCG) signals can yield useful information into the mechanics of the heart. A machine learning technique for detecting and classifying murmurs is proposed, which takes into account the statistically significant features derived from comparing spectrogram images obtained by the Short-Time Fourier Transform (STFT) of the PCG signals. The spectrograms are compared by Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Matrix (SSIM). Finally, these similarity index features are fed into various decision trees, both with and without PCA to classify like normal heart sound and murmurs like systolic, diastolic, and continuous. The SSIM and PSNR alone offer accuracy of 88.23% and 87.94%, respectively for distinguishing normal and murmur and are differ with a P-value of 2.05 × 10−19. The PCA enabled coarse tree performs better in terms of classification accuracy of 85% and 92.50% during training and testing, respectively. The results show that this method can accurately detect and classify heart murmurs, outperforming conventional methods. [ABSTRACT FROM AUTHOR]

Epilepsy seizure prediction is vital for enhancing the quality of life for individuals with epilepsy. In this study, we introduce a novel hybrid deep learning architecture, merging DenseNet and Vision Transformer (ViT) with an attention fusion layer for seizure prediction. DenseNet captures hierarchical features and ensures efficient parameter usage, while ViT offers self-attention mechanisms and global feature representation. The attention fusion layer effectively amalgamates features from both networks, guaranteeing the most relevant information is harnessed for seizure prediction. The raw EEG signals were preprocessed using the short-time Fourier transform (STFT) to implement time–frequency analysis and convert EEG signals into time–frequency matrices. Then, they were fed into the proposed hybrid DenseNet–ViT network model to achieve end-to-end seizure prediction. The CHB-MIT dataset, including data from 24 patients, was used for evaluation and the leave-one-out cross-validation method was utilized to evaluate the performance of the proposed model. Our results demonstrate superior performance in seizure prediction, exhibiting high accuracy and low redundancy, which suggests that combining DenseNet, ViT, and the attention mechanism can significantly enhance prediction capabilities and facilitate more precise therapeutic interventions. [ABSTRACT FROM AUTHOR]

Esophageal speech (ES) is a pathological voice that is often difficult to understand. Moreover, acquiring recordings of a patient's voice before a laryngectomy proves challenging, thereby complicating enhancing this kind of voice. That is why most supervised methods used to enhance ES are based on voice conversion, which uses healthy speaker targets, things that may not preserve the speaker's identity. Otherwise, unsupervised methods for ES are mostly based on traditional filters, which cannot alone beat this kind of noise, making the denoising process difficult. Also, these methods are known for producing musical artifacts. To address these issues, a self-supervised method based on the Only-Noisy-Training (ONT) model was applied, consisting of denoising a signal without needing a clean target. Four experiments were conducted using Deep Complex UNET (DCUNET) and Deep Complex UNET with Complex Two-Stage Transformer Module (DCUNET-cTSTM) for assessment. Both of these models are based on the ONT approach. Also, for comparison purposes and to calculate the evaluation metrics, the pre-trained VoiceFixer model was used to restore the clean wave files of esophageal speech. Even with the fact that ONT-based methods work better with noisy wave files, the results have proven that ES can be denoised without the need for clean targets, and hence, the speaker's identity is retained. [ABSTRACT FROM AUTHOR]

The microarray gene expression data poses a tremendous challenge due to their curse of dimensionality problem. The sheer volume of features far surpasses available samples, leading to overfitting and reduced classification accuracy. Thus the dimensionality of microarray gene expression data must be reduced with efficient feature extraction methods to reduce the volume of data and extract meaningful information to enhance the classification accuracy and interpretability. In this research, we discover the uniqueness of applying STFT (Short Term Fourier Transform), LASSO (Least Absolute Shrinkage and Selection Operator), and EHO (Elephant Herding Optimisation) for extracting significant features from lung cancer and reducing the dimensionality of the microarray gene expression database. The classification of lung cancer is performed using the following classifiers: Gaussian Mixture Model (GMM), Particle Swarm Optimization (PSO) with GMM, Detrended Fluctuation Analysis (DFA), Naive Bayes classifier (NBC), Firefly with GMM, Support Vector Machine with Radial Basis Kernel (SVM-RBF) and Flower Pollination Optimization (FPO) with GMM. The EHO feature extraction with the FPO-GMM classifier attained the highest accuracy in the range of 96.77, with an F1 score of 97.5, MCC of 0.92 and Kappa of 0.92. The reported results underline the significance of utilizing STFT, LASSO, and EHO for feature extraction in reducing the dimensionality of microarray gene expression data. These methodologies also help in improved and early diagnosis of lung cancer with enhanced classification accuracy and interpretability. [ABSTRACT FROM AUTHOR]

The spindle rotation error of computer numerical control (CNC) equipment directly reflects the machining quality of the workpiece and is a key indicator reflecting the performance and reliability of CNC equipment. Existing rotation error prediction methods do not consider the importance of different sensor data. This study developed an adaptive weighted deep residual network (ResNet) for predicting spindle rotation errors, thereby establishing accurate mapping between easily obtainable vibration information and difficult-to-obtain rotation errors. Firstly, multi-sensor data are collected by a vibration sensor, and Short-time Fourier Transform (STFT) is adopted to extract the feature information in the original data. Then, an adaptive feature recalibration unit with residual connection is constructed based on the attention weighting operation. By stacking multiple residual blocks and attention weighting units, the data of different channels are adaptively weighted to highlight important information and suppress redundancy information. The weight visualization results indicate that the adaptive weighted ResNet (AWResNet) can learn a set of weights for channel recalibration. The comparison results indicate that AWResNet has higher prediction accuracy than other deep learning models and can be used for spindle rotation error prediction. [ABSTRACT FROM AUTHOR]

For manufacturing components with thick plates, such as in the heavy equipment and shipbuilding industries, the gas metal arc welding (GMAW) process is applied. Among the components that apply the thick plate GMAW process, there are groove butt joints, which are fabricated through multi-pass welding. Various welding qualities are managed in multi-pass welding, and the root-pass weld is controlled to ensure complete joint penetration (CJP). Currently, the state of complete joint penetration during root-pass welding is managed visually, making it difficult to confirm the penetration condition in real time. Therefore, there is a need to predict the penetration condition in real time. In this study, we propose a convolutional neural network (CNN)-based prediction model that can classify penetration conditions using welding current and voltage data from the root pass of V-groove butt joints. The root gap of the joints was varied between 1.0 and 2.0 mm, and the wire feed rate was adjusted. During welding, the current and voltage were measured. The welding current and voltage are transformed into a short-time Fourier transform (STFT) representation depicting the arc and wire extension lengths. The transformed dynamic resistance STFT information serves as the input variable for the CNN model. Preprocessing steps, including thresholding, are applied to optimize the input variables. The CNN architecture comprises three convolutional layers and two pooling layers. The model classifies penetration conditions as partial joint penetration (PJP), CJP, and burn-through, achieving a high accuracy of 97.8%. The proposed method facilitates the non-destructive evaluation of the root-pass welding quality without expensive monitoring equipment, such as vision cameras. It is expected to be immediately applied to the thick plate welding process using readily available welding data. [ABSTRACT FROM AUTHOR]

A signal is primarily generated when two objects collide, with one falling from a specific height and the other disrupting its balance. This signal contains an important component that aids in the diagnosis of material damage. In this work, a piezoelectric thin film sensor made of polyvinylidene fluoride (PVDF) was employed to monitor the impact signal and process the recorded signal. By stacking a left and right piezoelectric film sensor, drop impact experiments were performed on rubber sheet material. The data-acquisition system captured the impact signal, and a signal processing tool was used to assess its performance. To determine the time-frequency domain visibility of the impact signal, the short-time Fourier transform (STFT) and continuous wavelet transform (CWT) were used. The detection of the elastic signal and impact force on the material is suitable for piezoelectric film. A steel impactor strikes a silicon rubber sheet at a speed of 3.43 m/s and exerts an average impact force of 72.6 kN. To make the impact signal more visible, the noise signal can be denoised using a bandpass filter. The continuous wavelet transform has a high time-frequency resolution compared to the short-time Fourier transform. Additionally, the analysis of non-stationary signals is improved by using STFT and CWT approaches. Thus, the elastic signal detection in the material may be recognized using the average total of the signals identified by the sensor. [ABSTRACT FROM AUTHOR]

Special tran function theory (STFT) is a powerful nonlinear problem-solving tool. In this paper, four different nonlinear power engineering problems in the field of induction machines, power inductors, perovskite solar cells, and supercapacitors are represented via the same transcendental equation. Furthermore, the analytical solution of the derived transcendental equation is expressed by using the STFT. Comparisons of the accuracy of the presented solutions with corresponding solutions determined with numerical calculation for all observed power engineering problems are also presented. It is shown that the proposed analytical solution is applicable, simple to implement, highly accurate and low-time consuming. Furthermore, in the mathematical sense, the structures of the final expressions for all observed variables in all observed problems are simpler than literature-known analytical solutions. The Mathematica codes for different STFT solutions are given as an appendix of this paper. [ABSTRACT FROM AUTHOR]

Diabetes is a chronic condition, which occurs due to felicitous regulation of glucose levels in the blood. In this work, time frequency distributions like STFT and CWT are explored to convert ECG signal into time frequency image (TFI). The GLCM features are extracted from TFI, these features are used as input to machine learning classifiers namely decision tree, naive Bayes, and support vector machine with different kernel functions. The features from CWT using the Fine Gaussian SVM classifier has provided an accuracy of 88.15%. The effectiveness of the proposed method provides by comparison of performance parameter for classification of diabetic and normal ECG signals. [ABSTRACT FROM AUTHOR]

This paper outlines a method for identifying and counteracting distributed denial of service (DDoS) and low-rate denial of service (DoS) attacks. These impair significant threats to network security and can disrupt the accessibility and efficacy of systems under attack. The proposed method combines Time-Frequency Analysis (TFA) using Short-Time Fourier Transform (STFT) and a Deep Learning model (DLM), namely Recurrent Neural Network (RNN), to enhance network security. By leveraging the strengths of STFT and RNN, the approach achieves improved detection capabilities and enables timely response and effective mitigation. The CICDDoS2019 dataset has been employed to conduct the evaluation, which provides a diverse set of realistic attack traffic scenarios. The results show that the proposed approach is effective, with an impressive accuracy rate of 99.1%. Compared to traditional methods, the integrated achieves higher accuracy and lower false positive rates. This research highlights the potential of Multimodal Fusion method, for addressing the growing need for advanced defense mechanisms in today's evolving threat landscape.

The influence of environmental noise is generally excluded during research on machine fault diagnosis using acoustic signals. This study proposes a fault diagnosis method using a variational autoencoder (VAE) and domain adaptation neural network (DANN), both of which are based on unsupervised learning, to address this problem. The proposed method minimizes the impact of environmental noise and maintains the fault diagnosis performance in altered environments. The fault diagnosis algorithm was implemented using acoustic signals containing noise, present in the malfunctioning industrial machine investigation and inspection open dataset, and the fault prediction performance in noisy environments was examined based on forklift acoustic data using the VAE and DANN. The VAE primarily learns from normal state acoustic data and determines the occurrence of faults based on reconstruction error. To achieve this, statistical features of Mel frequency cepstral coefficients were extracted, generating features applicable regardless of signal length. Additionally, features were enhanced by applying noise reduction techniques via magnitude spectral subtraction and feature optimization, reflecting the characteristics of rotating equipment. Furthermore, data were augmented using generative adversarial networks to prevent overfitting. Given that the forklift acoustic data possess time-series characteristics, the exponentially weighted moving average was determined to quantitatively track time-series changes and identify early signs of faults. The VAE defined the reconstruction error as the fault index, diagnosing the fault states and demonstrating excellent performance using time-series data. However, the fault diagnosis performance of the VAE tended to decrease in noisy environments. Moreover, applying DANN for fault diagnosis significantly improved diagnostic performance in noisy environments by overcoming environmental differences between the source and target domains. In particular, by adapting the model learned in the source domain to the target domain and considering the domain differences based on signal-to-noise ratio, high diagnostic accuracy was maintained regardless of the noise levels. The DANN evaluated interdomain similarity using cosine similarity, enabling the accurate classification of fault states in the target domain. Ultimately, the combination of the VAE and DANN techniques enabled effective fault diagnosis even in noisy environments.

Epilepsy is one of the most common brain disorders, and seizures of epilepsy have severe adverse effects on patients. Real-time epilepsy seizure detection using electroencephalography (EEG) signals is an important research area aimed at improving the diagnosis and treatment of epilepsy. This paper proposed a real-time approach based on EEG signal for detecting epilepsy seizures using the STFT and Google-net convolutional neural network (CNN). The CHB-MIT database was used to evaluate the performance, and received the results of 97.74 % in accuracy, 98.90 % in sensitivity, 1.94 % in false positive rate. Additionally, the proposed method was implemented in a real-time manner using the sliding window technique. The processing time of the proposed method just 0.02 s for every 2-s EEG episode and achieved average 9.85- second delay in each seizure onset.

The Visual Evoked Potential (VEP) examination is used to analyze the appropriate functioning of the optical pathways from the retina to the visual cortex. In hospitals, the diagnosis made by physicians is based mainly on reading the temporal trace and identifying the latency P100. However, after a considerable research effort, it has been confirmed that this method is subjective and relatively less reliable. In our work, we report different approaches to resolve the inadequacy of traditional classification, by studying the efficiency of VEP signal classification in a comparative approach using 3 models: Model A: STFT-CNN, Model B: CWT-CNN, and Model C: Wigner-Ville-CNN, therefore we evaluate in the same context the effectiveness of using a pre-trained 2D CNN structure. The time-frequency transformation allows us to generate two-dimensional data from one-dimensional signals to bring out the integrated features that are not valued in the temporal plot, and then exploit them for good discrimination between the two classes, in order to be able to use a CNN-2D classification architecture, taking into consideration the advantages offered by this architecture in terms of the involvement of the attribute extraction phase and its efficiency in classifying 2D data. The results provided by the different scenarios proved that the Wigner-Ville transformation combined with a pre-trained CNN architecture can be considered a good method in terms of different performance metrics, which demonstrates that it is a successful candidate for providing significant assistance to physicians in their analysis of VEP signals.

ABSTRAK Emosi memberikan peran penting dalam interaksi manusia yang didapat melalui respon yang tepat. Respon yang tak tepat menunjukan adanya gangguan mental sehingga diperlukan identifikasi emosi. Identifikasi dapat dilakukan menggunakan aktivitas sinyal listrik di otak menggunakan Elektroensephalogram (EEG). Karena sinyal EEG pada setiap kanal merupakan urutan data maka dijadikan multi-kanal yang direpresentasikan pada matriks agar urutan-urutan data tetap terjaga. Penggunaan matriks memadukan informasi dari ketiga dimensi (kanal x frekuensi x waktu) dapat menggambarkan kompleksitas dari sinyal EEG. Sehingga dapat mengenali pola aktivitas otak pada rentang frekuensi tertentu berkembang sepanjang waktu. Untuk menangkap informasi tersebut perlu dilakukan ekstraksi fitur agar mewakili variabel-variabel emosi. Ekstraksi dilakukan pada domain frekuensi (4-45 Hz) dan waktu menggunakan Short Time Fourier Transform (STFT) kemudian idenitifikasi menggunakan 3D Convolutional Neural Network (CNN). Eksperimen menggunakan 3D CNN menghasilkan akurasi 65.45 dengan teknik koreksi bobot Adamax. Kata kunci: emosi, sinyal EEG, multi-kanal, STFT, 3D-CNN ABSTRACT Emotions play an important role in human interaction through appropriate responses. Inappropriate responses indicate a mental disorder, so identification of emotions is required. Identification can be done using electrical signal activity in the brain with Electroencephalogram (EEG). Because the EEG signal in each channel is a data sequence, it is made into a multi-channel represented in a matrix so that the data sequence is maintained. Using a matrix combining information from all three dimensions (channel x frequency x time) can describe the complexity of the EEG signal. Allowing recognition of evolving brain activity patterns within specific frequency ranges over time. Extraction is done in the frequency domain (4-45 Hz) and time using Short Time Fourier Transform (STFT), then identification using a 3D Convolutional Neural Network (CNN). Experiments using 3D CNN resulted in an accuracy of 65.45 with the Adamax weight correction technique. Keywords: emotion, EEG signal, multi-channel, STFT, 3D-CNN

The rapid growth of Unmanned Aerial Vehicle (UAV) technology, or drone, has shown its popularity and has been significantly applied to various purposes today. Nevertheless, with all the sophistication of drones, many related topics are still attractive, especially when a drone is designed to carry out a cargo mission. Thus, in this research, the dynamic model of a Hexacopter drone to deliver goods belongs to PT Aero Terra Scan is being developed. This dynamic modeling aims to further the drone's development by modeling it in 2 cases: no-payload and with a payload of 5 kg cases. The dynamic model of this Hexacopter is based on flight dynamics, a field of science studied in Aeronautical Engineering, and is implemented using Simulink. Through the results of this research, several conclusions have been withdrawn: (1) The drone's unstable nature characteristic inherently, even though it is analyzed from the initial hover condition. Thus, the drone and its system as a whole can never be separated with the feedback control that made it can maneuver adequately. (2) Several technical parameters of this Hexacopter, including the geometry, mass, the moment of inertia, until the estimation of motor throttle is required to achieve its hover conditions, both in the no-payload case and with-payload of 5 kg case. (3) The Hexacopter basic dynamic system model is based on the flight dynamics until its motion system control tuning through root locus map analysis using Simulink.

Epilepsy seizure prediction is vital for enhancing the quality of life for individuals with epilepsy. In this study, we introduce a novel hybrid deep learning architecture, merging DenseNet and Vision Transformer (ViT) with an attention fusion layer for seizure prediction. DenseNet captures hierarchical features and ensures efficient parameter usage, while ViT offers self-attention mechanisms and global feature representation. The attention fusion layer effectively amalgamates features from both networks, guaranteeing the most relevant information is harnessed for seizure prediction. The raw EEG signals were preprocessed using the short-time Fourier transform (STFT) to implement time–frequency analysis and convert EEG signals into time–frequency matrices. Then, they were fed into the proposed hybrid DenseNet–ViT network model to achieve end-to-end seizure prediction. The CHB-MIT dataset, including data from 24 patients, was used for evaluation and the leave-one-out cross-validation method was utilized to evaluate the performance of the proposed model. Our results demonstrate superior performance in seizure prediction, exhibiting high accuracy and low redundancy, which suggests that combining DenseNet, ViT, and the attention mechanism can significantly enhance prediction capabilities and facilitate more precise therapeutic interventions.

Esophageal speech (ES) is a pathological voice that is often difficult to understand. Moreover, acquiring recordings of a patient’s voice before a laryngectomy proves challenging, thereby complicating enhancing this kind of voice. That is why most supervised methods used to enhance ES are based on voice conversion, which uses healthy speaker targets, things that may not preserve the speaker’s identity. Otherwise, unsupervised methods for ES are mostly based on traditional filters, which cannot alone beat this kind of noise, making the denoising process difficult. Also, these methods are known for producing musical artifacts. To address these issues, a self-supervised method based on the Only-Noisy-Training (ONT) model was applied, consisting of denoising a signal without needing a clean target. Four experiments were conducted using Deep Complex UNET (DCUNET) and Deep Complex UNET with Complex Two-Stage Transformer Module (DCUNET-cTSTM) for assessment. Both of these models are based on the ONT approach. Also, for comparison purposes and to calculate the evaluation metrics, the pre-trained VoiceFixer model was used to restore the clean wave files of esophageal speech. Even with the fact that ONT-based methods work better with noisy wave files, the results have proven that ES can be denoised without the need for clean targets, and hence, the speaker’s identity is retained.

Weak suction and large vibrations indicate an Induced Draft Fan (IDF) problem. The Fast Fourier Transform (FFT) method cannot be applied to non-stationary vibration signals. Therefore, this study aims to analyze non-stationary vibration signals for IDF vibration signals at start-up so that the source of damage to the IDF can be found. The research process begins with a brief measurement of both bearing locations with horizontal and axial axes. Processing of the vibration signal from the measurement using the FFT method and the Short Time Fourier Transform (STFT). Based on the STFT spectrogram graph for measurements on the horizontal and axial axes, the dominant frequency values are the same. The frequency with the largest amplitude value is at one RPM IDF or 25 Hz. High vibration at 1 RPM is a big indication that the IDF is experiencing unbalance.

For the problem of low time-frequency aggregation of the short-time Fourier transform (STFT), which causes the parameter estimation performance degradation of linear frequency modulation (LFM) signals at low signal-to-noise ratio (SNR), second-order synchrosqueezing transform (SSST) is proposed based on the square of STFT amplitude. The time-frequency resolution and energy aggregation are improved by means of squeezing and reassigning the time-frequency spectrum. Meanwhile, in order to decrease the calculation of classical parameter estimation methods, the Hough transform is used for rough estimation, and then the fractional Fourier transform (FRFT) is used for accuracy estimation based on the Renyi entropy. The simulation result shows that higher estimation accuracy is obtained at low SNR, and it has better robustness. [ABSTRACT FROM AUTHOR]

Featured Application: This paper proposes a novel method for transverse cracking detection based on vehicle vibration response. Transverse cracking is thought of as the typical distress of asphalt pavements. A faster detection technique can provide pavement performance information for maintenance administrations. This paper proposes a novel vehicle-vibration-based method for transverse cracking detection. A theoretical model of a vehicle-cracked pavement vibration system was constructed using the d'Alembert principle. A testing system installed with a vibration sensor was put in and applied to a testing road. Then, parameter optimization of the Short-time Fourier transform (STFT) was conducted. Transverse cracking and normal sections were processed by the optimized STFT algorithm, generating two ideal indicators. The maximum power spectral density and the relative power spectral density, which were extracted from 3D time–frequency maps, performed well. It was found that the power spectral density caused by transverse cracks was above 100 dB/Hz. The power spectral density at normal sections was below 80 dB/Hz. The distribution of the power spectral density for the cracked sections is more discrete than for normal sections. The classification model based on the above two indicators had an accuracy, true positive rate, and false positive rate of 94.96%, 92.86%, and 4.80%, respectively. The proposed vehicle-vibration-based method is capable of accurately detecting pavement transverse cracking. [ABSTRACT FROM AUTHOR]

Current deep learning-based speech enhancement methods focus on enhancing the time–frequency representation of the signal. However, conventional methods can lead to speech damage due to resolution mismatch problems that emphasize only specific information in the time or frequency domain. To address these challenges, this paper introduces a speech enhancement model designed with a dual-path structure that identifies key speech characteristics in both the time and time–frequency domains. Specifically, the time path aims to model semantic features hidden in the waveform, while the time–frequency path attempts to compensate for the spectral details via a spectral extension block. These two paths enhance temporal and spectral features via mask functions modeled as LSTM, respectively, offering a comprehensive approach to speech enhancement. Experimental results show that the proposed dual-path LSTM network consistently outperforms conventional single-domain speech enhancement methods in terms of speech quality and intelligibility. [ABSTRACT FROM AUTHOR]

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Background: The widely used rolling element bearings in rotating machines undergo progressive degradation with continuous operation. To identify bearing faults, complex time-frequency based signal processing techniques and high-end deep neural network algorithms have been used to perform fault classification, which is time-consuming. Method: In this paper, the focus was given to replace the complex time-frequency domain signal processing techniques by incorporating a simple time-domain based methodology. Initially, the vibration signature of different bearing faults was acquired at three different speeds and was directly converted into images by 2D-Vibration Imaging (2D-VI) technique using an overlapping-based moving window. The extracted images were fed into Convolutional Neural Network (CNN) for automatic feature extraction, followed by classification using Support Vector Machine (SVM). Results and validation: Separately, time-frequency spectrums were also extracted to compare the effectiveness of the proposed methodology. Furthermore, the proposed methodology was validated on the bearing dataset of combined faults and Case Western Reserve University (CWRU). Conclusion: The experimental results showed that the proposed methodology has the potential to replace the conventional approach by consuming less computational time without affecting classification accuracy. [ABSTRACT FROM AUTHOR]

For manufacturing components with thick plates, such as in the heavy equipment and shipbuilding industries, the gas metal arc welding (GMAW) process is applied. Among the components that apply the thick plate GMAW process, there are groove butt joints, which are fabricated through multi-pass welding. Various welding qualities are managed in multi-pass welding, and the root-pass weld is controlled to ensure complete joint penetration (CJP). Currently, the state of complete joint penetration during root-pass welding is managed visually, making it difficult to confirm the penetration condition in real time. Therefore, there is a need to predict the penetration condition in real time. In this study, we propose a convolutional neural network (CNN)-based prediction model that can classify penetration conditions using welding current and voltage data from the root pass of V-groove butt joints. The root gap of the joints was varied between 1.0 and 2.0 mm, and the wire feed rate was adjusted. During welding, the current and voltage were measured. The welding current and voltage are transformed into a short-time Fourier transform (STFT) representation depicting the arc and wire extension lengths. The transformed dynamic resistance STFT information serves as the input variable for the CNN model. Preprocessing steps, including thresholding, are applied to optimize the input variables. The CNN architecture comprises three convolutional layers and two pooling layers. The model classifies penetration conditions as partial joint penetration (PJP), CJP, and burn-through, achieving a high accuracy of 97.8%. The proposed method facilitates the non-destructive evaluation of the root-pass welding quality without expensive monitoring equipment, such as vision cameras. It is expected to be immediately applied to the thick plate welding process using readily available welding data.

Fault diagnosis and health monitoring of industrial rotating machines are of paramount importance for ensuring the reliability, safety, and efficiency of modern industrial operations. This paper proposes a Short-Time Fourier Transform (STFT)-based fault diagnosis approach for industrial rotating machinery. In this proposed model, the STFT of the reference vibration signals is evaluated and compared with the STFT of the other testing vibration signals to diagnose the fault types. Three different similarity operators: Euclidean distance, cosine similarity, and structural similarity are used to conclude the similarity index between the reference signal and test signal. By using variable speed vibration data with different fault types, the proposed model can better simulate real-world conditions and improve the accuracy and effectiveness of fault diagnosis. The results from the confusion matrices, heat maps, and t-SNE plots demonstrate the effectiveness of the proposed method for fault diagnosis and monitoring of variable-speed rotating machines using vibration signals. It is concluded that the structural similarity index proved to be a promising approach for accurate fault diagnosis in variable-speed rotating machines. The results are also compared with the existing approaches in the literature and it was concluded that the proposed model attains the highest accuracy for the variable speed rotating machines.

Motor imagery (MI) electroencephalography (EEG) technology is acquiring great attention from researchers due to its remarkable real-world applications. EEG signals inherit a high degree of non-stationarity, making their analysis not modest. Hence, choosing an appropriate signal processing approach becomes crucial. This comparative paper aims to identify a suitable signal processing method among famous approaches, namely short-time Fourier transform (STFT), continuous wavelet transform (CWT), and two variations of discrete wavelet transform maximal overlap DWT (MODWT) and MODWT multiresolution analysis (MODWTMRA). Different mother wavelet basis filters experimented with wavelet methods: Morse, Amor, Bump, Symlets, Daubechies, Coiflets, and Fejér-Korovkin. The different methods were tested on the classification of the right-hand and left-hand motor imagery tasks using the brain-computer interface (BCI) competition IV 2b dataset. A shallow convolutional neural network containing a single convolution layer was first trained and then used for classification. The experimental outcomes verified that MI EEG signals can be better analyzed and recognized using the maximal overlap-based signal processing methods. The classification accuracy proved that MODWT and MODWTMRA with the Symlets wavelet outperformed the other methods.

Modulated signal recognition is difficult but essential for applications like cognitive radio, intelligent communication, radio supervision, and electronic countermeasure. Current modulation recognition models lack comprehensiveness and typicality of various signals and primarily rely on artificial feature extraction. In this study, a convolutional neural network (CNN)-based method for modulated signal recognition is proposed. The proposed method converts modulation recognition into image identification. To increase the acuity of CNN for learning time-frequency features, channel attention and spatial attention are further introduced based on the fused features. Eight different types of modulated signals, including Rect, LFM, Barker, GFSK, CPFSK, B-FM, DSB-AM, and SSB-AM, are used in the experiments. The recognition rate of the proposed model is greater than 85% when the SNR (signal-to-noise ratio) is greater than -10dB, and it ranges from 92% to 98% when the SNR is 0dB. The recognition rate of the proposed method outperforms the two other comparison methods, CNN without an attention mechanism and LSTM. [ABSTRACT FROM AUTHOR]