Article
Chemistry, Multidisciplinary
Yiqin Shao, Jiale Tang, Jun Liu, Lixin Han, Shijian Dong
Summary: This paper proposes a method to intelligently predict key output variables that are difficult to measure online in complex systems, by using the LASSO algorithm, TCN-LSTM network, and SAM mechanism for dynamic modeling. The accuracy of sequence prediction is further improved by selecting the principal component variables and optimizing the network.
Article
Engineering, Multidisciplinary
Ganapathi Ammasai Sengodan
Summary: This work presents a novel method to predict the mechanical responses of arbitrary microstructures using deep learning, generating two-phase microstructural images, quantifying them using the two-point statistical homogenisation scheme, and projecting microstructures and stress-strain data into lower order orthogonal spaces by principal component analysis to minimize computational efforts. By using convolutional neural networks to learn reduced order statistically homogeneous microstructures and stress-strain data, the study successfully predicts the mechanical responses of randomly generated two-phase fibre reinforced plastic (FRP) composite microstructures with better accuracy and minimal computational effort.
COMPOSITES PART B-ENGINEERING
(2021)
Article
Automation & Control Systems
Libiao Bai, Chaopeng Song, Xinyu Zhou, Yuanyuan Tian, Lan Wei
Summary: The purpose of this study is to develop a project portfolio risk (PPR) assessment model using an enhanced backpropagation neural network. The model considers project interdependencies and uses fuzzy logic and principal component analysis (PCA) for data processing. An improved genetic algorithm (IGA) is used to optimize the initial weights and thresholds of the neural network. The results show that the established model performs better in assessing PPR compared to other methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Chemistry, Analytical
Michele Alessandrini, Giorgio Biagetti, Paolo Crippa, Laura Falaschetti, Simona Luzzi, Claudio Turchetti
Summary: This study aims to develop an automatic classification method that can effectively handle corrupted EEG data, by utilizing a recurrent neural network and preprocessing with robust principal component analysis. The results demonstrate that even with data corruption, RPCA can successfully filter outlying data components, improving detection accuracy.
Article
Computer Science, Information Systems
Shashi Bhushan, Mohammed Alshehri, Neha Agarwal, Ismail Keshta, Jitendra Rajpurohit, Ahed Abugabah
Summary: Recognizing facial expressions is a major challenge, especially in the field of industrial Internet of Things. This paper proposes a new framework, FRS-DCT-SVM, that utilizes discrete cosine transform and genetic algorithm optimization for face detection and feature extraction. The framework achieves better results in terms of clustering time and accuracy compared to other methods.
Article
Engineering, Mechanical
Mahmoud Elsamanty, Abdelkader Ibrahim, Wael Saady Salman
Summary: This study proposes a method based on data fusion to generate new fused signatures for the diagnosis of mechanical faults in rotating machines. The weighted decision fusion method combines the outputs of multiple sensors to generate the fused decision. Principal Component Analysis (PCA) is applied to reduce the dimensionality of evaluating parameters and a backpropagation neural network (BPNN) is constructed for fault diagnosis. The results show that the proposed method achieves superior data fusion solutions and PCA in the condition monitoring of rotating machines.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Geochemistry & Geophysics
Giuseppe Guarino, Matteo Ciotola, Gemine Vivone, Giovanni Poggi, Giuseppe Scarpa
Summary: This paper proposes a simple yet effective method to adapt unsupervised CNNs from MS to HS image fusion. By using PCA-based decorrelation transform, a significant portion of HS image energy is compressed into few bands. A suitably adapted pansharpening network is then used to super-resolve the principcal components. Experimental results show high performance in both quantitative and qualitative evaluations, outperforming state-of-the-art methods.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Chemistry, Multidisciplinary
Geraud Bossavi, Rongguo Yan, Muhammad Irfan
Summary: In this study, a deep learning algorithm was developed to accurately calculate blood pressure and heart rate by extracting features from PPG, ECG, and ABP signals. The proposed approach achieved better results and lower errors compared to other methods, making it significant in the healthcare field.
APPLIED SCIENCES-BASEL
(2023)
Article
Agronomy
Ya Tian, Limin Xie, Mingyang Wu, Biyun Yang, Captoline Ishimwe, Dapeng Ye, Haiyong Weng
Summary: This study evaluated the potential of multicolor fluorescence imaging combined with PCA and SVM for early detection of salt stress in plants. By analyzing Arabidopsis with this method, it was proven that multicolor fluorescence imaging can effectively differentiate control and salt-stressed plants, providing an important tool for monitoring and studying plant stress.
Article
Agronomy
Shahrzad Zolfagharnassab, Abdul Rashid Bin Mohamed Shariff, Reza Ehsani, Hawa Ze Jaafar, Ishak Bin Aris
Summary: The research aims to develop a thermal imaging method to accurately indicate the maturity of oil palm fruits. By testing 297 samples, the results show that Delta Temp can reliably classify oil palm FFBs. The study also finds that the ANN method achieves the highest classification accuracy.
Article
Green & Sustainable Science & Technology
Meiyan Li, Yingjun Fu
Summary: Supply Chain Finance (SCF) is a new type of financing business carried out by commercial banks based on supply chain management. This paper proposes a supply chain financial credit risk prediction model based on PCA-GA-SVM, which can effectively improve the credit risk management ability of Supply Chain Finance.
Article
Instruments & Instrumentation
Xin Wang, Ailun Zhang, Sheng Liang, Shuqin Lou
Summary: In order to reduce the nuisance alarm rate in phase-sensitive OTDR sensing system, a novel event identification model based on PCA and PNN was proposed. By training this model, five kinds of disturbance events can be effectively identified with an average identification rate of 97.74% and an average response time of 0.93 s. The high identification rate and fast response time make this method more adaptable in practical application.
INFRARED PHYSICS & TECHNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Jili Tao, Zheng Yu, Ridong Zhang, Furong Gao
Summary: Neural network prediction and data processing are widely used in the chemical industry, but the traditional models have poor accuracy due to disturbance variables. This paper proposes a method that includes PCA dimensionality reduction, RBF neural network model, LM algorithm, and GA to improve modeling accuracy. The method shows significant improvement in prediction accuracy compared to other RBF neural network modeling methods.
APPLIED SOFT COMPUTING
(2021)
Article
Engineering, Electrical & Electronic
Dan Zhang, Yongyi Chen, Fanghong Guo, Hamid Reza Karimi, Hui Dong, Qi Xuan
Summary: This article presents a new intelligent fault diagnosis method for rolling bearings using convolutional neural network (CNN) and fuzzy C-means (FCM) clustering algorithm. The proposed method can automatically extract features from the vibration signals of rolling bearings and identify different fault types. The results indicate that this method achieves higher accuracy compared to existing literature.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Hardware & Architecture
R. Tamilarasi, S. Prabu
Summary: Hyperspectral imagery is useful for determining urban-related characteristics such as roads, trees, buildings, and structures. Researchers are currently focusing on deep learning and machine learning methods for image classification. This research proposes a new technique for dimensionality reduction and classification, combining ICA, PCA, FCN, and SVM models, to extract road and building features with high accuracy from hyperspectral images. Experimental results show better accuracy compared to existing machine learning approaches.
JOURNAL OF SUPERCOMPUTING
(2021)
Article
Mathematics
Naheed Akhtar, Mubbashar Saddique, Khurshid Asghar, Usama Ijaz Bajwa, Muhammad Hussain, Zulfiqar Habib
Summary: This paper provides a detailed review of existing passive video tampering detection techniques and analyzes the state-of-the-art research work and commonly used datasets. The limitations of existing algorithms are discussed, and future research challenges and directions are proposed.
Article
Computer Science, Information Systems
Zulfiqar Ali, Fazal-e Amin, Muhammad Hussain
Summary: This study proposes a novel hybrid approach using fragile zero watermarking to protect the privacy and integrity of patients' personal information and medical data transmitted from remote health facilities. By incorporating visual cryptography and chaotic randomness, the algorithm effectively prevents information leakage.
Article
Mathematics
Muhammad Nadeem Ashraf, Muhammad Hussain, Zulfiqar Habib
Summary: Diabetic retinopathy (DR) is a vision-threatening complication, and a deep convolutional neural network (CNN) can effectively diagnose and screen DR patients. Training deep models with minimal data is challenging, but fine-tuning pre-trained CNNs and making architectural amendments can improve performance. The modified model (DR-ResNet50) outperforms state-of-the-art methods in terms of various metrics and shows high sensitivity and low false-positive rate in testing, demonstrating its value and suitability for early screening.
Article
Mathematics
Fahman Saeed, Muhammad Hussain, Hatim A. Aboalsamh
Summary: Fingerprint recognition systems often face the challenge of high processing complexity, especially when multiple sensors are used. In this study, a technique for automatically determining the architecture of a CNN model adaptive to fingerprint classification was proposed, which effectively helps design lightweight models and enhances the speed of the recognition process. The method was evaluated using benchmark datasets and outperformed existing models and techniques.
Article
Mathematics
Ebtihal Al-Mansour, Muhammad Hussain, Hatim A. Aboalsamh, Fazal-e-Amin
Summary: Masses are early indicators of breast cancer, and distinguishing between benign and malignant masses is a challenging problem. Existing deep learning methods for mass classification on mammograms have shown unsatisfactory performance, but a new method introduced in this study has outperformed the state-of-the-art approaches on public benchmark datasets.
Article
Computer Science, Artificial Intelligence
Rawan Alsughayer, Muhammad Hussain, Fahman Saeed, Hatim AboalSamh
Summary: Remote sensing image tampering detection is crucial for hiding important information, but existing methods lack robustness. In this study, we propose an image-to-image transformation method based on U-Net architecture, focusing on residual noise and introducing a constrained convolutional layer. The method effectively detects and localizes splicing tampering.
APPLIED INTELLIGENCE
(2023)
Article
Chemistry, Analytical
Ashwaq Alotaibi, Muhammad Hussain, Hatim AboAlSamh, Wadood Abdul, George Bebis
Summary: This study addresses the interoperability issue of fingerprint sensors by proposing a method that uses deep learning to enhance fingerprints captured by different sensor types. Through training with edge loss and the combination of two learning frameworks, the proposed method achieves effective fingerprint quality enhancement on benchmark datasets, outperforming existing methods.
Article
Mathematics
Mariam Busaleh, Muhammad Hussain, Hatim A. A. Aboalsamh, Fazal-e-Amin, Sarah A. Al Sultan
Summary: This paper proposes TwoViewDensityNet, a deep learning-based method for mammographic breast density classification. By using two different views for breast classification, this method achieves high overall performance on public datasets.
Article
Mathematics
Fahman Saeed, Muhammad Hussain, Hatim A. A. Aboalsamh, Fadwa Al Adel, Adi Mohammed Al Owaifeer
Summary: Diabetic retinopathy (DR) is a leading cause of blindness in middle-aged diabetic patients, and regular screening using fundus imaging is crucial to detect complications and delay disease progression. To address the time-consuming and subjective nature of manual screening, a method for automatically customizing CNN models based on fundus image lesions is proposed, which outperforms existing pre-trained CNN models and other neural architecture search models.
Article
Mathematics
Amani Alahmadi, Muhammad Hussain, Hatim Aboalsamh
Summary: This paper proposes a deep CNN periocular recognition model called LDA-CNN, which incorporates an LDA layer after the last convolution layer of a pre-trained CNN backbone model. The LDA layer enforces the model to learn discriminative features with small within-class variation and large between-class separation. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods for periocular recognition in unconstrained environments and generalizes well to different conditions.
Article
Mathematics, Applied
Walaa Alsumari, Muhammad Hussain, Laila Alshehri, Hatim A. Aboalsamh
Summary: Using biometric modalities for person recognition is crucial, and recent studies have focused on using electroencephalography (EEG) as a secure modality. Existing methods have limitations in terms of usability and generalization, but this study addresses these issues by proposing a lightweight convolutional neural network (CNN) model. The proposed EEG-based recognition system achieved high performance on a benchmark dataset, with a rank-1 identification result of 99% and an equal error rate of 0.187%.
Article
Medicine, General & Internal
Farah Muhammad, Muhammad Hussain, Hatim Aboalsamh
Summary: A multimodal emotion recognition method based on deep canonical correlation analysis (DCCA) is proposed in this study, which fuses electroencephalography (EEG) and facial video clips. The proposed method achieves an average accuracy of 93.86% and 91.54% on MAHNOB-HCI and DEAP datasets, respectively.
Article
Mathematics
Hend Alshaya, Muhammad Hussain
Summary: Accurately identifying seizure types is crucial for treating epilepsy patients. This paper presents a deep network model based on ResNet and LSTM for classifying seizure types from EEG trials. The proposed model outperforms other state-of-the-art models with an F1-score of 97.4%.
Article
Computer Science, Information Systems
Muhammad Hussain, Emad-Ul-Haq Qazi, Hatim A. Aboalsamh, Ihsan Ullah
Summary: This study proposes an automatic emotion recognition system based on deep learning and electroencephalogram signals. It introduces a lightweight pyramidal one-dimensional convolutional neural network model with a small number of learnable parameters, and a two-level ensemble classifier. The method scans each channel incrementally in the first level and fuses the predictions using majority vote. In the second level, predictions from all channels are fused to predict the emotional state. The method was validated using the DEAP dataset and achieved high accuracies in distinguishing different emotion states.
Article
Computer Science, Information Systems
Emad-ul-Haq Qazi, Muhammad Hussain, Hatim A. AboAlsamh
Summary: The study proposed an automatic alcoholism detection system using deep learning technology, utilizing a multi-channel Pyramidal neural convolutional network. By analyzing EEG signals, the system can accurately detect whether the subject is alcoholic or normal, demonstrating robustness and effectiveness on the KDD dataset.
CMC-COMPUTERS MATERIALS & CONTINUA
(2021)