Convolutional Neural Network Structure Research Articles

Deep Learning Model for Cosmetic Gel Classification Based on a Short-Time Fourier Transform and Spectrogram.

Cosmetics and topical medications, such as gels, foams, creams, and lotions, are viscoelastic substances that are applied to the skin or mucous membranes. The human perception of these materials is complex and involves multiple sensory modalities. Traditional panel-based sensory evaluations have limitations due to individual differences in sensory receptors and factors such as age, race, and gender. Therefore, this study proposes a deep-learning-based method for systematically analyzing and effectively identifying the physical properties of cosmetic gels. Time-series friction signals generated by rubbing the gels were measured. These signals were preprocessed through short-time Fourier transform (STFT) and continuous wavelet transform (CWT), respectively, and the frequency factors that change over time were distinguished and analyzed. The deep learning model employed a ResNet-based convolution neural network (CNN) structure with optimization achieved through a learning rate scheduler. The optimized STFT-based 2D CNN model outperforms the CWT-based 2D and 1D CNN models. The optimized STFT-based 2D CNN model also demonstrated robustness and reliability through k-fold cross-validation. This study suggests the potential for an innovative approach to replace traditional expert panel evaluations and objectively assess the user experience of cosmetics.

ОДНОВИМІРНА ЗГОРТКОВА МОДЕЛЬ НЕЙРОННОЇ МЕРЕЖІ ДЛЯ ОБРОБЛЕННЯ СИГНАЛІВ З АМПЛІТУДНОЮ МОДУЛЯЦІЄЮ БАГАТЬОХ СКЛАДОВИХ

The processing of radio signals using artificial neural networks (ANNs) has great potential for research, which can be explained by the adaptability of ANNs to various transmission conditions and the ability to detect abstract patterns of changes in signal parameters. The article reviews the works of other authors devoted to different ways of using ANNs for processing radio signals. Taking into account the information in the reviewed works, the research task was formed, which consists in developing an optimized ANN model for radio signal processing. Signals with amplitude modulation of many components (AMMC) were chosen to form training samples for ANN. The choice of modulation type is justified by greater energy efficiency compared to other widely used digital modulation types, such as quadrature amplitude modulation. Mathematic basis of AMMC signal generation is described. The process of finding the coordinates of three component 8-AMMC signal constellation is explained, the formation of signals in the time plane based on the found coordinates is explained as well as their discretization and the addition of white noise. An iterative algorithm for generating initial data for ANN based on the described ratios is proposed. The general structure of one-dimensional convolutional neural network is considered. Functions of individual neurons, connections between them, the formation of layers and the convolution operation are described mathematically. On the basis of the previously given ratios, a final display of the network was formed. Specific dimensions and activation functions for layers are selected. The use of convolutional layers is justified by time invariance. Based on the reviewed mathematical models, selected activation functions and dimensions, a neural model was formed. The process of validating the effectiveness of the formed neural model is described, which is based on comparing the symbolic error probabilities of the proposed and reference models at different signal-to-noise ratios. The validation results are presented. The advantages of the obtained model over the previously proposed purely recurrent model and the AMMC reference receiver are explained.

Improving the Accuracy of Direction of Arrival Estimation with Multiple Signal Inputs Using Deep Learning

In this paper, an innovative cyclic noise reduction method and an improved CAPON algorithm (also the called minimum variance distortionless response (MVDR) algorithm) are proposed to improve the accuracy and reliability of DOA (direction of arrival) estimation. By processing the eigenvalues obtained from the covariance matrix of the received signal, the signal-to-noise ratio (SNR) can be increased by up to 5 dB by the cyclic noise reduction method, which will improve the DOA estimation accuracy. The improved CAPON algorithm has a convolution neural network (CNN) structure, whose input is the processed covariance matrix of the received signal, and the CAPON spectral value is used as the training label to obtain the estimated spatial spectrum. It retains the advantages of the CAPON algorithm, which can achieve blind source estimation, and simulations show that the proposed algorithm exhibits a better performance than the traditional algorithm in conditions of various SNRs and snapshot numbers. The simulation results show that, based on a certain SNR, the root mean square error (RMSE) of the improved CAPON algorithm can be reduced from 0.86° to 0.8° compared to traditional algorithms, and the angle estimation error can be decreased by up to about 0.3°. With the help of the cyclic noise reduction method, the angle estimation error decreases from 0.04° to 0.02°.

Global Receptive Field Designed Complex-Valued Convolutional Neural Network Equalizer for Optical Fiber Communications

In this paper, an improved complex-valued convolutional neural network (CvCNN) structure to be placed at the received side is proposed for nonlinearity compensation in a coherent optical system. This complex-valued global convolutional kernel-assisted convolutional neural network equalizer (CvGNN) has been verified in terms of Q-factor performance and complexity compared to seven other related nonlinear equalizers based on both the 64 QAM experimental platform and the QPSK numerical platform. The global convolution operation of the proposed CvGNN is more suitable for the calculation process of perturbation coefficients, and the global receptive field can also be more effective at extracting effective information from perturbation feature maps. The introduction of CvCNN can directly focus on the complex-valued perturbation feature maps themselves without separately processing the real and imaginary parts, which is more in line with the waveform-dependent physical characteristics of optical signals. Based on the experimental platform, compared with the real-valued neural network with small convolutional kernel (RvCNNC), the proposed CvGNNC improves the Q-factor by ∼2.95 dB at the optimal transmission power, while reducing the time complexity by ∼44.7%.

Large-scale land use/land cover extraction from Landsat imagery using feature relationships matrix based deep-shallow learning

Deep learning has demonstrated its effectiveness in capturing high-level features, with convolutional neural networks (CNNs) excelling in remote sensing classification. However, CNNs encounter challenges when applied to Landsat images with limited multi-spectral bands, as they struggle to learn stable features from the spectral domain and integrate them with spatial features to enhance accuracy. Additionally, most CNN applications focus on learning features directly from the raw image, making them susceptible to spectral environment changes. To overcome these limitations, this paper introduces a novel approach for large-scale Land Use/Land Cover (LULC) extraction from Landsat OLI images. The proposed classification architecture comprises two modules. The first module utilizes a feature relationships matrix to generate an extent feature map (EFM), and a specifically designed CNN structure learns deep features from the EFM and spatial domain. In the second module, a multiple classifiers system (MCS) is employed to obtain shallow learning features, which are further enhanced by another CNN structure through continued learning. The combined features from these modules contribute to improved classification of remote sensing images. Experimental results demonstrate that our proposed method effectively acquires stable features for training deep learning models with strong generalization ability. It exhibits significant advancements in accuracy improvement and large-scale LULC extraction in the Yangtze River Economic Belt (YREB) in China, surpassing comparative approaches based on deep learning and non-deep learning methods.

Detection and Classification of Rolling Bearing Defects Using Direct Signal Processing with Deep Convolutional Neural Network

Currently, great emphasis is being placed on the electrification of means of transportation, including aviation. The use of electric motors reduces operating and maintenance costs. Electric motors are subjected to various types of damage during operation, of which rolling bearing defects are statistically the most common. This article focuses on presenting a diagnostic tool for bearing conditions based on mechanic vibration signals using convolutional neural networks (CNN). This article presents an alternative to the well-known classical diagnostic tools based on advanced signal processing methods such as the short-time Fourier transform, the Hilbert–Huang transform, etc. The approach described in the article provides fault detection and classification in less than 0.03 s. The proposed structures achieved a classification accuracy of 99.8% on the test set. Special attention was paid to the process of optimizing the CNN structure to achieve the highest possible accuracy with the fewest number of network parameters.

Application of continuous wavelet transform and convolutional neural networks in fault diagnosis of PMSM stator windings

Efficiency, reliability, and durability play a key role in modern drive systems in line with the Industry 4.0 paradigm and the sustainability trend. To ensure this, highly efficient motors and appropriate systems must be deployed to monitor their condition and diagnose faults during the operation. For these reasons, in recent years, more and more research has been focused on developing new methods for fault diagnosis of permanent magnet synchronous motors (PMSMs). This paper proposes a novel hybrid method for the automatic detection and classification of PMSM stator winding faults based on combining the continuous wavelet transform (CWT) analysis of the negative sequence component of the stator phase currents with a convolutional neural network (CNN). CWT scalogram images are used as the inputs of the CNN-based interturn short circuits fault classifier model. Experimental tests were carried out to verify the effectiveness of the proposed approach under various motor operating conditions and at an incipient stage of fault propagation. In addition, the effects of the input image format, CNN structure, and training process parameters on model accuracy and classification effectiveness were investigated. The results of the experimental tests confirmed the high effectiveness of fault detection (99.4%) and classification (97.5%), as well as other important advantages of the developed method.

Applications of the TL-Based Fault Diagnostic System for the Capacitor in Hybrid Aircraft

The article concerns the problem of capacitor diagnosis in a hybrid aircraft. Capacitors are one of the most commonly damaged components of electrical vehicle drive systems. The result of these failures is an increase in voltage ripple. Most known analytical methods are based on frequency spectrum analysis, which is time-consuming and computationally complex. The use of deep neural networks (DNNs) allows for the direct use of the measurement signal, which reduces the operating time of the overall diagnostic system. However, the problem with these networks is the long training process. Therefore, this article uses transfer learning (TL), which allows for the secondary use of previously learnt DNNs. To collect data to learn the network, a test bench with the ability to simulate a capacitor failure was constructed, and a model based on it was made in the MATLAB/Simulink environment. A convolutional neural network (CNN) structure was developed and trained by the TL method to estimate the capacitance of the capacitor based on signals from the Simulink-designed model. The proposed fault diagnostic method is characterised by a nearly 100% efficiency in determining capacitance, with an operating time of about 10 ms, regardless of load and supply voltage.

Open-circuit fault diagnosis method of inverter in wind turbine yaw system based on GADF image coding

Considering the concealment of inverter open-circuit faults in yaw system of wind turbines, a method for diagnosing open-circuit fault of yaw system inverter is proposed based on Gramian Angular Difference Field (GADF) image coding in this article. Firstly, the current vector phase in the vicinity of relatively high yaw speed is collected as monitoring data. Secondly, considering the insufficient fault feature extraction ability of Convolutional Neural Network (CNN) structures with single dimensional data input during model training, the one-dimensional current vector phase is encoded into two-dimensional image by GADF image coding and an effective CNN fault diagnosis model is obtained with fewer fault samples. Finally, by comparing with actual monitoring data of wind turbine, the effectiveness of yaw system simulation model is verified. The proposed method can achieve identification and localization of open-circuit fault for single and two power devices, which is found to have better anti-noise interference effect and the results are not affected by yaw angle and mechanical torque. It is approved that the proposed method is effective with a simulation on the RT-LAB platform.

Positive incentive CNN structure coupled nonconvex model for image super-resolution

Abstract This paper studies super-resolution (SR) technique to reconstruct high-quality images for deep image analysis. Currently, the convolutional neural networks (CNNs) are well performing methods and the finding that random noise added in the network can have positive incentive effect, we innovatively propose a positive incentive CNNs. However, concerning the uncontrollable characteristic and lack consistency of deep network, we propose a novel framework that joins nonconvex model based on framelet and positive incentive CNN structure, which can impose consistency between the high-resolved image and the given low-resolution image, and depict image information by sparse representation. Furthermore, to overcome the challenge of computing the minimizer of the nonconvex problem, we use proximal linearized minimization (PLM) algorithm to convex the nonconvex term, then apply the alternating direction method of multipliers (ADMM) as the solver which can converge to a stationary point of the nonconvex model. The experimental outcomes on Set5, Set14, BSD100, Urban100, and real-world images demonstrate that the proposed approach outperforms the state-of-the-art methods in terms of peak signal to noise ratio (PSNR) value, structural similarity index (SSIM), and visual quality.

ARTIFICIAL INTELLIGENCE APPLIED TO QUALITY DEFECTS CLASSIFICATION FOR TYRES

The manufacture of radial tires is complex, and for this reason some defects inevitably end up appearing in the production process. Currently the commonly used defect detection is manual detection, among these techniques are ultrasound testing, thermal detection, X-ray detection, digital holography, among others. The X-ray or technical radiography allows the analysis of all the internal components of the tire, such as the bead structure and the different layers of rubber and cords, whether metallic or textile. The task of detecting defects using X-ray images is done manually, which causes loss of time and costs for the company. Furthermore, it is a subjective, inefficient, time-consuming and even biased process as it requires a high level of concentration and focus. The objective of this work is the digitization of the conventional process of analysis of quality failures in metallic radial tires using algorithms developed with Artificial Intelligence resources, capable of identifying and classifying defects present in X-Ray images. This work presents a brief introduction about what the tire is, as well as the development of the proposed algorithm, using classic image segmentation techniques with a Computer Vision library and two different convolutional neural network structures. Through the comparison, it was verified that the algorithm based on U-Net obtained more relevant results, however, the algorithm based on Mask R-CNN was also promising and could be useful in new works, with a larger database and different settings parameters.

A serial semantic segmentation model based on encoder-decoder architecture

The thriving progress of Convolutional Neural Networks (CNNs) and the outstanding efficacy of Visual Transformers (ViTs) have delivered impressive outcomes in the domain of semantic segmentation. However, each model in isolation entails a trade-off between high computational complexity and compromised computational efficiency. To address this challenge, we effectively combine the CNN and encoder-decoder structures in a Transformer-inspired fashion, presenting the Serial Semantic Segmentation Trans via CNN Former (SSS-Former) model. To augment the feature extraction capability, we utilize the meticulously crafted SSS-CSPNet, resulting in a well-designed architecture for the holistic model. We propose a novel SSS-PN attention network that enhances the spatial topological connections of features, leading to improved overall performance. Additionally, the integration of SASPP bridges the semantic gap between multi-scale features and enhances segmentation ability for overlapping objects. To fulfill the requirement of real-time segmentation, we leverage a novel restructuring technique to devise a more lightweight and faster ResSSS-Former model. Abundant experimental results demonstrate that both SSS-Former and ResSSS-Former outperform existing state-of-the-art methods in terms of computational efficiency, result precision, and speed. Remarkably, SSS-Former achieves a mIoU of 58.63 % at 89.1FPS on the ADE20K dataset. On the validation and testing datasets of CityScapes, it obtains mIoU scores of 85.1 % and 85.2 % respectively, with a speed of 94.1FPS. Our optimized ResSSS-Former achieves impressive real-time segmentation results, with an astonishing 100+FPS while maintaining high segmentation accuracy. The compelling results from the ISPRS datasets further validate the effectiveness of our proposed models in segmenting multi-scale and overlapping objects.

An Effective Image Classification Method for Plant Diseases with Improved Channel Attention Mechanism aECAnet Based on Deep Learning

Since plant diseases occurring during the growth process are a significant factor leading to the decline in both yield and quality, the classification and detection of plant leaf diseases, followed by timely prevention and control measures, are crucial for safeguarding plant productivity and quality. As the traditional convolutional neural network structure cannot effectively recognize similar plant leaf diseases, in order to more accurately identify the diseases on plant leaves, this paper proposes an effective plant disease image recognition method aECA-ResNet34. This method is based on ResNet34, and in the first and the last layers of this network, respectively, we add this paper’s improved aECAnet with the symmetric structure. aECA-ResNet34 is compared with different plant disease classification models on the peanut dataset constructed in this paper and the open-source PlantVillage dataset. The experimental results show that the aECA-ResNet34 model proposed in this paper has higher accuracy, better performance, and better robustness. The results show that the aECA-ResNet34 model proposed in this paper is able to recognize diseases of multiple plant leaves very accurately.

Hindustani raga and singer classification using 2D and 3D pose estimation from video recordings

Using pose estimation with video recordings, we apply an action recognition machine learning algorithm to demonstrate the use of the movement information to classify singers and the ragas (melodic modes) they perform. Movement information is derived from a specially recorded video dataset of solo Hindustani (North Indian) raga recordings by three professional singers each performing the same nine ragas, a smaller duo dataset (one singer with tabla accompaniment) as well as recordings of concert performances by the same singers. Data is extracted using pose estimation algorithms, both 2D (OpenPose) and 3D. A two-pathway convolutional neural network structure is proposed for skeleton action recognition to train a model to classify 12-second clips by singer and raga. The model is capable of distinguishing the three singers on the basis of movement information alone. For each singer, it is capable of distinguishing between the nine ragas with a mean accuracy of 38.2% (with the most successful model). The model trained on solo recordings also proved effective at classifying duo and concert recordings. These findings are consistent with the view that while the gesturing of Indian singers is idiosyncratic, it remains tightly linked to patterns of melodic movement: indeed we show that in some cases different ragas are distinguishable on the basis of movement information alone. A series of technical challenges are identified and addressed, with code shared alongside audiovisual data to accompany the paper.

Transfer learning achieves high recall for object classification in fluvial environments with limited data

Field surveys to collect data from fluvial ecosystems traditionally focus on specific phenomena related to geomorphology or hydrology. Low-cost unmanned aerial vehicles (UAVs) additionally empower the fast and massive collection of airborne photogrammetry, providing geospatially explicit information. This remote sensing data complements field surveys by offering contextual information on geomorphological conditions, including digital terrain models. AI-based image recognition can augment contextual information to extrapolate archetypal object classes through name labels, such as “gravel”, “sand”, “plant”, or “large wood”. However, obtaining sufficient ground truth data for these classifications, particularly in morphodynamic fluvial environments, is challenging and induces high costs. This study introduces a transfer learning approach to address the challenge of low data availability, enabling AI-based mapping of complex objects in fluvial landscapes. We leverage the learned general structure of a deep convolutional neural network (CNN) pre-trained on a broad range of images. The fixed latent features of the pre-trained CNN stem from GoogLeNet. A fixed feature extractor serves to classify objects with limited data amounts. Satisfactory performance is measured with a recall rate, expressing the ability of a model to find all occurrences of a class on an image. High spatial heterogeneity in the locations of measurements on the x-y plane improves model performance. With a minimum of 400 labeled instances, the model achieves a satisfactory 93.75-% recall for a “large wood” target class, providing evidence of the effectiveness of transfer learning in remote sensing for geomorphological studies. This ability to detect large woods in river environments is critical to restoration efforts as it helps create fish habitat, which is essential to supporting biodiversity.

INCREASING ROBUSTNESS OF I-VECTORS VIA MASKING: A CASE STUDY IN SYNTHETIC SPEECH DETECTION

Ensuring security in speaker recognition systems is crucial. In the past years, it has been demonstrated that spoofing attacks can fool these systems. In order to deal with this issue, spoof speech detection systems have been developed. While these systems have served with a good performance, their effectiveness tends to degrade under noise. Traditional speech enhancement methods are not efficient for improving performance, they even make it worse. In this research paper, performance of the noise mask obtained via a convolutional neural network structure for reducing the noise effects was investigated. The mask is used to suppress noisy regions of spectrograms in order to extract robust i-vectors. The proposed system is tested on the ASVspoof 2015 database with three different noise types and accomplished superior performance compared to the traditional systems. However, there is a loss of performance in noise types that are not encountered during training phase.

A Data-Driven Autonomous Assessment Framework for Education Quality Based on Multiscale Deep Learning

Nowadays, computational intelligence-assisted autonomous assessment of education quality has become a more and more general concern in the area of smart education management. As education quality assessment is a complicated process with multiple heterogeneous factors, it remains challenging to make effective assessment using simple information modality and criteria. To deal with this issue, this paper introduces multiscale deep learning, and proposes a novel data-driven autonomous assessment framework for education quality. In particular, four aspects of heterogeneous indicators are selected as the basic indexes and a dilated convolutional neural network structure is formulated to perform multiscale feature extraction. Then, the structural equation is adopted to make multivariate characterization and output the final assessment results. At last, some simulations are carried out on realistic education operation data to evaluate the efficiency of the proposed autonomous assessment framework. Two aspects of findings can be deduced from the results. For one thing, multisource fusion of multiple indicators well makes sense to autonomous education assessment. For another, multiscale deep learning can provide some beneficial promotion to assessment efficiency.

Research on spacecraft in orbit perception based on artificial neural networks and digital twin technology using grating arrays

In order to improve the safety of spacecraft, the research on artificial neural network and digital twin technology based on, to our best knowledge, a novel fiber Bragg grating (FBG) sensor array is proposed for intelligent sensing monitoring of spacecraft on-orbit collisions. Femtosecond FBG arrays were fabricated on the novel oxide-doped fiber by point-by-point writing technique. The femtosecond FBG is analyzed using the time-dependent perturbation theory of quantum mechanics. The FBG array can achieve high-temperature measurement of 1100 °C and large strain measurement of 15000 µε. The sensing arrays were deployed on the surface of the spacecraft. Constructed the multi-layer perceptron neural network structure and convolutional neural network structure. 1200 samples were trained. Conducted model accuracy testing. The accuracy rate is above 98%, and accuracy verification has been implemented. The digital twin model was designed based on various data such as strain and temperature of the spacecraft structure under impact monitored by FBG sensors. A precise mapping has been formed between the physical entities of spacecraft and digital twins. Empower spacecraft with functions such as self-monitoring, judgment, and response. To ensure the stable and safe operation of spacecraft.

A review of convolutional neural networks in computer vision

In computer vision, a series of exemplary advances have been made in several areas involving image classification, semantic segmentation, object detection, and image super-resolution reconstruction with the rapid development of deep convolutional neural network (CNN). The CNN has superior features for autonomous learning and expression, and feature extraction from original input data can be realized by means of training CNN models that match practical applications. Due to the rapid progress in deep learning technology, the structure of CNN is becoming more and more complex and diverse. Consequently, it gradually replaces the traditional machine learning methods. This paper presents an elementary understanding of CNN components and their functions, including input layers, convolution layers, pooling layers, activation functions, batch normalization, dropout, fully connected layers, and output layers. On this basis, this paper gives a comprehensive overview of the past and current research status of the applications of CNN models in computer vision fields, e.g., image classification, object detection, and video prediction. In addition, we summarize the challenges and solutions of the deep CNN, and future research directions are also discussed.

Rapid quantitative analysis of cotton-polyester blended fabrics using near-infrared spectroscopy combined with CNN-LSTM

Accurate and rapid quantitative analysis of fiber content is important for blended fabric quality control and supervision. In this paper, a deep learning method based on an improved hybrid network of convolutional neural network (CNN) and long short-term memory (LSTM) to fast and accurate determine cotton content of cotton-polyester blended fabrics using near-infrared (NIR) spectroscopy. The cotton-polyester blended fabric samples were partitioned into training set, validation set and testing set in the ratio of 6:2:2 to develop the quantitative model of cotton content of blended fabric. Firstly, a one-dimensional CNN architecture is designed and modeled to obtain the CNN parameters and to evaluate the spectral preprocessing methods. Then, the CNN structure is modified and improved using LSTM to obtain a CNN-LSTM structure. The root mean square error (RMSE) of the validation set obtained by CNN model is 0.83%, which is 34.03% lower than that of the traditional partial least squares (PLS) model, and the RMSE of the testing set obtained by CNN model is 0.88%, which is 25.88% lower than that of the PLS model. The RMSE of the validation set obtained by CNN-LSTM model is 0.65%, which is 47.85% lower than that of the PLS model, and the RMSE of the testing set obtained by CNN-LSTM model is 0.75%, which is 36.57% lower than that of the PLS model. Compared with PLS and CNN models, the CNN-LSTM model can more accurately analyze the cotton content proportion of cotton-polyester blended fabrics, which provides a new fast and non-invasive method for the quantitative detection of fabrics.