Recent Submissions

  • Cancelable template generation based on quantization concepts

    Rana M. Nassar; Ashraf A. M. Khalaf; Ghada M. El-Banby; Fathi E. Abd El-Samie; Hussein, Aziza; Walid El-Shafai; External Collaboration; NA; NA; NA; et al. (2023-06-28)
    The idea of cancelable biometrics is widely used nowadays for user authentication. It is based on encrypted or intentionally-distorted templates. These templates can be used for user verification, while keeping the original user biometrics safe. Multiple biometric traits can be used to enhance the security level. These traits can be merged together for cancelable template generation. In this paper, a new system for cancelable template generation is presented depending on discrete cosine transform (DCT) merging and joint photographic experts group (JPEG) compression concepts. The DCT has an energy compaction property. The low-frequency quartile in the DCT domain maintains most of the image energy. Hence, the first quartile from each of the four biometrics for the same user is kept and other quartiles are removed. All kept coefficients from the four biometric images are concatenated to formulate a single template. The JPEG compression of this single template with a high compression ratio induces some intended distortion in the template. Hence, it can be used as a cancelable template for the user acquired from his four biometric traits. It can be changed according to the arrangement of biometric quartiles and the compression ratio used. The proposed system has been tested through merging of face, palmprint, iris, and fingerprint images. It achieves a high user verification accuracy of up to 100%. It is also robust in the presence of noise.
  • Design and implementation of a low-cost circuit for medium-speed flash analog to digital conversions

    Nashaat M. Hussain Hassan; Mohamed Adel Esmaeel Salama; Hussein, Aziza; Mohamed Mourad Mabrook; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; et al. (2024-01-09)
    Despite the considerable advancements in analog-to-digital conversion (ADC) circuits, many papers neglect several crucial considerations: Firstly, it does not ensure that ADCs work well in the software or hardware. Secondly, it is not certain that ADCs have a wide range of amplitude responses for the input voltages to be convenient in many applications, especially in electronics, communications, computer vision, CubeSat circuits, and subsystems. Finally, many of these ADCs need to look at the suitability of the proposed circuit to the most extensive range of frequencies. In this paper, a design of a low-cost circuit is proposed for medium-speed flash ADCs. The proposed circuit is simulated based on a set of electronic components with specific values to achieve high stability operation for a wide range of frequencies and voltages, whether in software or hardware. This circuit is practically implemented and experimentally tested. The proposed design aims to achieve high efficiency in the sampling process over a range of amplitudes from 10 mV to 10 V. The proposed circuit operates at a bandwidth of frequencies from 0 Hz to greater than 10 kHz in the simulation and hardware implementation.
  • Efficient implementation of double random phase encoding and empirical mode decomposition for cancelable biometrics

    Gerges M. Salama; Walid El-Shafai; Safaa El-Gazar; Basma Omar; A. A. Hassan; Hussein, Aziza; Fathi E. Abd El-Samie; External Collaboration; NA; NA; et al. (2023-10-19)
    Biometric-based systems for secure access to different services have gained a significant attention in recent years. To ensure the protection of biometric data from potential hackers, it is crucial to store them in the form of secure templates. Cancelable templates offer an effective solution through allowing template replacement in case of security breaches. In this paper, we propose a novel unimodal cancelable biometric system that works on bio-signals such as voiceprint, electroencephalography (EEG), and electrocardiography (ECG) signals. The key feature of our proposed system is the utilization of Empirical Mode Decomposition (EMD) to decompose the bio-signals into different Intrinsic Mode Functions (IMFs). Among these IMFs, the first IMF, which carries the majority of the signal energy and distinguishes the bio-signal, plays a pivotal role in our system. To ensure the security of the cancelable biometric template, an encryption algorithm is employed. We use the Double Random Phase Encoding (DRPE) algorithm along with its random phase masks to encrypt the first IMF after converting it into 2D format. The use of DRPE and its random masks ensures a non-invertible transformation, which enhances the security of the encrypted data. To generate the cancelable template, we replace the first IMF of a reference signal with the encrypted first IMF obtained from the bio-signal. The resulting template retains the essential distinguishing characteristics of the bio-signal, while safeguarding its security. The verification process in our system involves matching of the encrypted first IMF of the stored templates with the encrypted first IMF of a new input signal. Extensive simulation analysis has been conducted to evaluate the performance of the proposed system. Various metrics, including Equal Error Rate (EER) and Area under Receiver Operating Characteristic curve (AROC), have been considered. The results demonstrate the high performance and stability of our system, even in the presence of different levels of white Gaussian noise, with an EER close to 0 and an AROC close to 1. In conclusion, our work presents an efficient implementation of the DRPE and EMD for the development of a robust and secure cancelable biometric system. The proposed system shows promising results and holds great potential for enhancing the security and reliability of biometric-based access control.
  • Enhancing Leukemia Detection: An Automated Approach Using Deep Learning and Ensemble Techniques

    Saad Ahmed Syed; Humaira Nisar; Rabeea Jaffari; Yan Chai Hum; Lee Yu Jen; Mian Qaisar, Saeed; External Collaboration; NA; NA; NA; et al. (Elsevier, 2024-01-11)
    As leukemia ranks high among the global causes of death, it's crucial to identify it early to enhance the prognosis for patients. The majority of diagnostic processes used today rely on medical experts inspecting samples manually. This is a laborious process that lacks an automated detection mechanism and takes a lot of time. With a focus on acute lymphoblastic leukemia (ALL), this work suggests an automated diagnostic method that uses Deep Learning (DL)-based ensembles to improve leukemia detection and prediction. We propose to utilize a combination of ten DL techniques (ResNet, ResNeXt, SE-ResNet, Inception V3, VGG, and its variants) and three ensemble techniques (Max voting, Averaging, and Stacking) to constitute the leukemia detection models and observe their performances. The ALL IDB benchmark leukemia dataset was evaluated using these techniques, with performances measured across several metrics namely: classification accuracy, F1 score, precision, recall (sensitivity), Kappa index, and ROC-AUC score. The findings from the experiments demonstrate a notable enhancement in leukemia detection performance when utilizing the proposed techniques. In particular, the proposed Ensemble Max Voting technique surpasses all other stateof-the-art detection models in the literature with an accuracy of 100.0% and an F1 score of 0.997. The main achievement of this study is the identification of the most effective method among various models and techniques for detecting leukemia.
  • Satellite Imagery-Based Cloud Classification Using Deep Learning

    Rukhsar Yousaf; Hafiz Zia Rehman; Khurram Jadoon; Zeashan H. Khan; Adnan Fazil; Zahid Mahmood; Mian Qaisar, Saeed; Abdul Jabbar Siddiqui; External Collaboration; NA; et al. (MDPI, 2023-12-01)
    A significant amount of satellite imaging data is now easily available due to the continued development of remote sensing (RS) technology. Enabling the successful application of RS in real-world settings requires efficient and scalable solutions to extend their use in multidisciplinary areas. The goal of quick analysis and precise classification in Remote Sensing Imaging (RSI) is often accomplished by utilizing approaches based on deep Convolution Neural Networks (CNNs). This research offers a unique snapshot-based residual network (SnapResNet) that consists of fully connected layers (FC-1024), batch normalization (BN), L2 regularization, dropout layers, dense layer, and data augmentation. Architectural changes overcome the inter-class similarity problem while data augmentation resolves the problem of imbalanced classes. Moreover, the snapshot ensemble technique is utilized to prevent over-fitting, thereby further improving the network’s performance. The proposed SnapResNet152 model employs the most challenging Large-Scale Cloud Images Dataset for Meteorology Research (LSCIDMR), having 10 classes with thousands of high-resolution images and classifying them into respective classes. The developed model outperforms the existing deep learning-based algorithms (e.g., AlexNet, VGG-19, ResNet101, and EfficientNet) and achieves an overall accuracy of 97.25%.
  • A computational study of a laminar methane–air flame assisted by nanosecond repetitively pulsed discharges

    Xiao Shao; Kabbaj, Narjisse; Deanna A Lacoste; Hong G Im; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; et al. (IOP Science, 2024-02-19)
    Nanosecond repetitively pulsed (NRP) discharges have been considered a promising technique for enhancing combustion efficiency and control. For successful implementation, it is necessary to understand the complex plasma–combustion interactions involving chemical, thermal, and hydrodynamic pathways. This paper aims to investigate the mechanisms enhancing a laminar methane–air flame assisted by NRP discharges by high fidelity simulations of the jet-wall burner employed in a previous experimental study. A phenomenological plasma model is used to represent the plasma energy deposition in two channels: (1) the ultrafast heating and dissociation of $\mathrm{O_2}$ resulting from the relaxation of electronically excited $\mathrm{N_2}$, and (2) slow gas heating stemming from the relaxation of $\mathrm{N_2}$ vibrational states. The flame displacement, key radical distribution and flame response under plasma actuation are compared with experimental results in good agreement. The computational model allows a systematic investigation of the dominant physical mechanism by isolating different pathways. It is found that the kinetic effect from atomic O production dominates the flame dynamics, while the thermal effect plays a minor role. Hydrodynamic perturbations arising from weak shock wave propagation appear to be sensitive to burner geometry and is found to be less significant in the case under study.
  • An Efficient Approach for the Detection and Prevention of Gray-Hole Attacks in VANETs

    Malik, Abdul; Khan, Muhammad Zahid; Mian Qaisar, Saeed; Faisal, Mohammad; Mehmood, Gulzar; External Collaboration; Artificial Intelligence & Cyber Security Lab; 0; 0; Electrical and Computer Engineering; et al. (IEEE, 2023-09-15)
    Vehicular Ad-Hoc Networks (VANETs) deliver a wide range of commercial as well as safety applications and further motivate the advancements of Internet of Vehicles (IoV), Intelligent Transportation Systems (ITS), and Vehicles to Everything (V2X) communication. Despite their potential benefits, VANETs are susceptible to a variety of security attacks due to their open, distributed, and dynamic nature, which includes intrinsic protocol design issues. One such an infamous security attack is the Gray-Hole Attack (GHA), typically has two variants: Smart GHA and Sequence Number-based GHA. In Smart GHA, the malicious node behaves normally during the route discovery process, while in Sequence Number-based GHA, the malicious node starts misbehaving during the route discovery process. In either case, once the route is successfully established, it starts dropping the packets. In this paper, a novel security approach called ‘‘Detection and Prevention of GHA’’ (DPGHA) is proposed to detect and prevent both variants of GHA in Ad Hoc On-Demand Distance Vector (AODV) based VANETs. The approach is based on the generation of dynamic threshold values of abnormal differences of received, forwarded, and generated control or data packets among nodes and their sequence numbers. The proposed DPGHA is implemented and tested in NS-2 and SUMO simulators and its various performances are compared with the most relevant benchmark approaches. The results showed that the proposed DPGHA performed better than the benchmark approaches in terms of reduced routing overhead by 10.85% and end-to-end delay by 3.85%, increased Packet Delivery Ratio (PDR) by 4.67% and throughput by 6.58%, and achieved a maximum detection rate of 2.3%.
  • Investigating the Optimal DOD and Battery Technology for Hybrid Energy Generation Models in Cement Industry Using HOMER Pro

    Basheer, Yasir; Mian Qaisar, Saeed; Waqar, Asad; Lateef, Fahad; Alzahrani, Ahmad; External Collaboration; Energy Lab; 0; 0; Electrical and Computer Engineering; et al. (IEEE, 2023-10-01)
    The cement industry is a major energy consumer, with most of its costs associated with fuel and energy requirements. While traditional thermal power plants generate electricity, they are both harmful and inefficient. In this study, battery depth of discharge (DOD) is evaluated for four different battery technologies in the context of the cement industry. The battery technologies evaluated are lead-acid (LA), lithium-ion (Li-ion), vanadium redox (VR), and nickel-iron (Ni-Fe). Five cement plants in Pakistan are considered, including Askari Cement Plant, Wah (ACPW), Bestway Cement Plant, Kalar Kahar (BCPKK), Bestway Cement Plant, Farooqia (BCPF), Bestway Cement Plant, Hattar (BCPH), and DG Cement Plant, Chakwal (DGCPC). Four hybrid energy generation models (HEGMs) were proposed using the HOMER pro software. HEGM-1 combines a diesel generator, photovoltaic system, converter, and battery system, while HEGM-2 consists of a photovoltaic system, converter, and battery system. HEGM-3 is a grid-connected version of HEGM-1 and HEGM-4 is the grid-connected version of HEGM-2. A reference base model using only grid connection is also considered. A multi-criteria decision analysis (MCDA) was performed using a cumulative objective function (COF) that includes net present cost (NPC), levelized cost of energy (LCOE), and greenhouse gas (GHG) emissions. The main objective was to maximize COF while minimizing NPC, LCOE, and GHG emissions using optimal battery technology and DOD. The results indicate that VR is the most optimal battery technology, with a DOD of 10% achieved in DGCPC using HEGM-3. This results in a 61.49% reduction in NPC, 78.62% reduction in LCOE, and 84.00% reduction in GHG emissions compared to the base model.
  • ECG based apnea detection by multirate processing hybrid of wavelet-empirical decomposition Hjorth features extraction and neural networks

    Khandelwal, Khandelwal; Salankar, Nilima; Mian Qaisar, Saeed; Upadhyay, Jyoti; Pławiak, Paweł; External Collaboration; Biometrics and Sensory Systems Lab; 0; 0; Electrical and Computer Engineering; et al. (Plos, 2023-11-02)
    Sleep Apnea (SA) can cause health complications including heart stroke and neurological disorders. The Polysomnography (PSG) test can detect the severity of sleep disturbance. However, it is expensive and requires a dedicated sleep laboratory and expertise to examine the patients. Therefore, it is not available to a large population in developing countries. This leads to the development of cost-effective and automated patient examination methods for the detection of sleep apnea. This study suggests an approach of using the ECG signals to categorize sleep apnea. In this work, we have devised an original technique of feature space designing by intelligently hybridizing the multirate processing, a mix of wavelet-empirical mode decomposition (W-EMD), modes-based Hjorth features extraction, and Adam-based optimized Multilayer perceptron neural network (MLPNN) for automated categorization of apnea. A publicly available ECG dataset is used for evaluating the performance of the suggested approach. Experiments are performed for four different sub-bands of the considered ECG signals. For each selected sub-band, five "Intrinsic Mode Functions" (IMFs) are extracted. Onward, three Hjorth features: complexity, activity, and mobility are mined from each IMF. In this way, four feature sets are formed based on wavelet-driven selected sub-bands. The performance of optimized MLPNN, for the apnea categorization, is compared for each feature set. Five different evaluation parameters are used to assess the performance. For the same dataset, a systematic comparison with current state-of-the-artwork has been done. Results have shown a classification accuracy of 98.12%.
  • A Novel Integration Technique for Optimal Location & Sizing of DG Units With Reconfiguration in Radial Distribution Networks Considering Reliability

    Raza, Ali; Zahid, Muhammad; Chen, Jinfu; Mian Qaisar, Saeed; Ilahi, Tehseen; Waqar, Asad; Alzahrani, Ahmad; External Collaboration; Energy Lab; 0; et al. (IEEE, 2023-11-02)
    This paper introduces an advanced approach for optimizing the distribution network reconfiguration (DNR) with the placement and sizing of multiple types of distributed generators (DGs). The method employs the ant colony optimization algorithm (ACOA), which is an innovative adaptive optimization algorithm, while also considering the system’s reliability. The primary objectives of the optimization problem are to minimize active power loss (APL), reduce voltage drop ( $V_{D}$ ) on buses, enhance system stability (SS), and improve overall reliability by reducing energy not supplied (ENS) to end-users. The optimization process involves determining the optimal location and size of DGs in the radial distribution network (RDN) using the ACOA meta-heuristic. The method maintains the radial structure of the system by selectively opening lines during the DNR process. The proposed technique is evaluated through simulations carried out on the IEEE-33 & -69 bus RDNs under various scenarios. The optimal solution is achieved by combining DG Type-1 with integration of DNR to reduce the APL and amplify the $V_{p}$ of buses in both RDNs. In this scenario APL is reduced to 87.97% (IEEE-33) and 92.83% (IEEE-69), respectively. Similarly, the $V_{p}$ of the buses significantly improved to 0.9776 p.u. (IEEE-33) and 0.9888 p.u. (IEEE-69), respectively. The results demonstrate the superiority of the presented ACOA-based approach over other techniques, such as fireworks algorithm (FWA) and adaptive shuffled frogs leaping algorithm (ASFLA). Combining the DNR and DGs placement in a simultaneous manner yields the best performance for the distribution network, resulting in lower APL, reduced $V_{D}$ , improved SS, and enhanced reliability. Furthermore, considering reliability in the optimization process significantly reduces ENS for customers and enables meeting their maximum load demand. Overall, the concurrent consideration of DNR and DGs placement using ACOA proves to be more effective than alternative algorithms.
  • EEG-based schizophrenia classification using penalized sequential dictionary learning in the context of mobile healthcare

    Haider, Usman; Hanif, Muhammad; Rashid, Ahmer; Mian Qaisar, Saeed; Subasi, Abdulhamit; External Collaboration; Biometrics and Sensory Systems Lab; 0; 0; Electrical and Computer Engineering; et al. (Elsevier, 2024-04-12)
    Mobile healthcare is an appealing approach based on the Internet of Medical Things (IoMT) and cloud computing. It can lead to unobstructed, economical, and patient-centric healthcare solutions. The key performance indicators of such systems are dimensionality reduction, computational effectiveness, low latency, and accuracy. In this context, a novel approach is devised for EEG-based schizophrenia, a severe mental disorder that adversely affects a person’s behavior and classification. A multichannel EEG recording with suitable granularity is required for precise analysis. It can increase exponentially the data dimensionality plus complexity and computational load. The proposed solution attains an interesting trade-off between dimensionality reduction plus computational effectiveness versus accuracy. It uses the penalized sequential dictionary learning (PSDL) that incorporates channel selection. First, PSDL learns a dictionary from the input data and evaluates its performance on all EEG channels. Based on this evaluation, a subset of six channels is selected for further training in the dictionary. The proposed PSDL algorithm then incorporates a penalty term that enhances the power of the learned dictionary on the selected channels. We evaluate the proposed approach on the multi-channel EEG dataset from the Institute of Psychiatry and Neurology in Warsaw, Poland. A performance comparison is also made with counterparts. The models’ performance depends on the EEG signals’ complexity. Therefore, we tried to make our models robust and straightforward, achieving appropriate performance with minimal computational cost. The proposed method reduces the dimension in two steps. First, the count of channels is reduced to 68.42%. In the second step, the kept information, 31.58% of channels, is further reduced to 83.75% using dictionary learning. The proposed framework secures a remarkable data dimension reduction and a lower computational cost and latency than the counterparts while attaining the sparse representation classification accuracy of 89.12%. These findings are promising and confirm the potential of investing in incorporating the proposed method in contemporary mobile healthcare solutions.
  • EEG-based emotion recognition using modified covariance and ensemble classifiers

    Subasi, Abdulhamit; Mian Qaisar, Saeed; College collaboration; Artificial Intelligence & Cyber Security Lab; 0; 0; Electrical and Computer Engineering; 0; Subasi, Abdulhamit (Springer, 2023-11-01)
    The Electroencephalography (EEG)-based precise emotion identification is one of the most challenging tasks in pattern recognition. In this paper, an innovative EEG signal processing method is devised for an automated emotion identification. The Symlets-4 filters based “Multi Scale Principal Component Analysis” (MSPCA) is used to denoise and reduce the raw signal’s dimension. Onward, the “Modified Covariance” (MCOV) is used as a feature extractor. In the classification step, the ensemble classifiers are used. The proposed method achieved 99.6% classification accuracy by using the ensemble of Bagging and Random Forest (RF). It confirms effectiveness of the devised method in EEG-based emotion recognition.
  • Investigation of vibration’s effect on driver in optimal motion cueing algorithm

    Hazoor, Ahmad; Tariq, Muhammad; Yasin, Awais; Razzaq, Sohail; Ahmad Chaudhry, Muhammad; Shaikh, Inam Ul Hasan; Ali, Ahsan; Mian Qaisar, Saeed; Iqbal, Jamshed; External Collaboration; et al. (Plos, 2023-11-30)
    The increased sensation error between the surroundings and the driver is a major problem in driving simulators, resulting in unrealistic motion cues. Intelligent control schemes have to be developed to provide realistic motion cues to the driver. The driver’s body model incorporates the effects of vibrations on the driver’s health, comfort, perception, and motion sickness, and most of the current research on motion cueing has not considered these factors. This article proposes a novel optimal motion cueing algorithm that utilizes the driver’s body model in conjunction with the driver’s perception model to minimize the sensation error. Moreover, this article employs H∞ control in place of the linear quadratic regulator to optimize the quadratic cost function of sensation error. As compared to state of the art, we achieve decreased sensation error in terms of small root-mean-square difference (70%, 61%, and 84% decrease in case of longitudinal acceleration, lateral acceleration, and yaw velocity, respectively) and improved coefficient of cross-correlation (3% and 1% increase in case of longitudinal and lateral acceleration, respectively).
  • Response of one-dimensional ionised layer to oscillatory electric fields

    Kabbaj, Narjisse; Hong G. Im; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; Kabbaj, Narjisse (2023-01-17)
    To provide fundamental insights into the response of laminar flames to alternating current (AC) electric fields, a simplified one-dimensional model using an ionised layer model is formulated with the conservation equations for the ion species with ionisation, recombination, and transport due to molecular diffusion and electric mobility. A parametric study is conducted to investigate the response of the ion layer at different voltages and oscillation frequencies, and the results are examined mainly in terms of the net current–voltage (I–V) characteristics. As the oscillation frequency is increased, a nonmonotonic response in the I–V curve is seen such that the current may exceed the saturation condition corresponding to the steady DC condition. In general the current reaches a peak as the unsteady time scale becomes comparable to the ion transport time scale, which is dictated by the mobility, and eventually becomes attenuated at higher frequencies to behave like a low-pass filter. The extent of the peak current rise and the cut-off frequency are found to depend on the characteristic time scales of the ion chemistry and mobility-induced transport. The simplified model serves as a framework to characterise the behaviour of complex flames in terms of the dominant ionisation and transport processes. The current overshoot behaviour may also imply that the overall effect of the electric field may be further magnified under the AC conditions, motivating further studies of multi-dimensional flames for the ionic wind effects.
  • Interoperable IoMT Approach for Remote Diagnosis with Privacy-Preservation Perspective in Edge Systems

    Subramaniam, E.V.D.; Srinivasan, K.; Mian Qaisar, Saeed; Pławiak, P.; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; et al. (MDPI, 2023-08-28)
    The emergence of the Internet of Medical Things (IoMT) has brought together developers from the Industrial Internet of Things (IIoT) and healthcare providers to enable remote patient diagnosis and treatment using mobile-device-collected data. However, the utilization of traditional AI systems raises concerns about patient privacy. To address this issue, we present a privacy-enhanced approach for illness diagnosis within the IoMT framework. Our proposed interoperable IoMT implementation focuses on optimizing IoT network performance, including throughput, energy consumption, latency, packet delivery ratio, and network longevity. We achieve these improvements using techniques such as device authentication, energy-efficient clustering, environmental monitoring using Circular-based Hidden Markov Model (C-HMM), data verification using Awad’s Entropy-based Ten-Fold Cross Entropy Verification (TCEV), and data confidentiality using Twine-LiteNet-based encryption. We employ the Search and Rescue Optimization algorithm (SRO) for optimal route selection, and the encrypted data are securely stored in a cloud server. With extensive network simulations using ns-3, our approach demonstrates substantial enhancements in the specified performance metrics compared with previous works. Specifically, we observe a 20% increase in throughput, a 15% reduction in packet drop rate (PDR), a 35% improvement in network lifetime, and a 10% decrease in energy consumption and delay. These findings underscore the efficacy of our approach in enhancing IoT network interoperability and protection, fostering improved patient care and diagnostic capabilities.
  • Automated Categorization of Multiclass Welding Defects Using the X-ray Image Augmentation and Convolutional Neural Network

    Mian Qaisar, Saeed; Dalila Say; Salah Zidi; Krichen Moez; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; et al. (MDPI, 2023-07-14)
    The detection of weld defects by using X-rays is an important task in the industry. It requires trained specialists with the expertise to conduct a timely inspection, which is costly and cumbersome. Moreover, the process can be erroneous due to fatigue and lack of concentration. In this context, this study proposes an automated approach to identify multi-class welding defects by processing the X-ray images. It is realized by an intelligent hybridization of the data augmentation techniques and convolutional neural network (CNN). The proposed data augmentation mainly performs random rotation, shearing, zooming, brightness adjustment, and horizontal flips on the intended images. This augmentation is beneficial for the realization of a generalized trained CNN model, which can process the multi-class dataset for the identification of welding defects. The effectiveness of the proposed method is confirmed by testing its performance in processing an industrial dataset. The intended dataset contains 4479 X-ray images and belongs to six groups: cavity, cracks, inclusion slag, lack of fusion, shape defects, and normal defects. The devised technique achieved an average accuracy of 92%. This indicates that the approach is promising and can be used in contemporary solutions for the automated detection and categorization of welding defects.
  • Causal speech enhancement using dynamical-weighted loss and attention encoder-decoder recurrent neural network

    Salem, Nema; Peracha, Fahad Khalil; Irfan Khattak, Muhammad; Saleem, Nasir; External Collaboration; NA; 0; 0; Electrical and Computer Engineering; Muhammad Irfan Khattak; et al. (2023-05-11)
    Speech enhancement (SE) reduces background noise signals in target speech and is applied at the front end in various real-world applications, including robust ASRs and real time processing in mobile phone communications. SE systems are commonly integrated into mobile phones to increase quality and intelligibility. As a result, a low-latency system is required to operate in real-world applications. On the other hand, these systems need effi cient optimization. This research focuses on the single-microphone SE operating in real time systems with better optimization. We propose a causal data-driven model that uses attention encoder-decoder long short-term memory (LSTM) to estimate the time-frequency mask from a noisy speech in order to make a clean speech for real-time applications that need low-latency causal processing. The encoder-decoder LSTM and a causal attention mechanism are used in the proposed model. Furthermore, a dynamical-weighted (DW) loss function is proposed to improve model learning by varying the weight loss values. Experi ments demonstrated that the proposed model consistently improves voice quality, intelligi bility, and noise suppression. In the causal processing mode, the LSTM-based estimated suppression time-frequency mask outperforms the baseline model for unseen noise types. The proposed SE improved the STOI by 2.64% (baseline LSTM-IRM), 6.6% (LSTM-KF), 4.18% (DeepXi-KF), and 3.58% (DeepResGRU-KF). In addition, we examine word error rates (WERs) using Google’s Automatic Speech Recognition (ASR). The ASR results show that error rates decreased from 46.33% (noisy signals) to 13.11% (proposed) 15.73% (LSTM), and 14.97% (LSTM-KF).
  • Hybrid EEG-fNIRS brain-computer interface based on the non-linear features extraction and stacking ensemble learning

    Asmaa Maher; Mian Qaisar, Saeed; N. Salankar; Feng Jiang; Ryszard Tadeusiewicz; Paweł Pławiak; Ahmed A. Abd El-Latif; Mohamed Hammad; External Collaboration; NA; et al. (Elsevier, 2023-05-31)
    The Brain-computer interface (BCI) is used to enhance the human capabilities. The hybrid-BCI (hBCI) is a novel concept for subtly hybridizing multiple monitoring schemes to maximize the advantages of each while minimizing the drawbacks of individual methods. Recently, researchers have started focusing on the Electroencephalogram (EEG) and “Functional Near-Infrared Spectroscopy” (fNIRS) based hBCI. The main reason is due to the development of artificial intelligence (AI) algorithms such as machine learning approaches to better process the brain signals. An original EEG-fNIRS based hBCI system is devised by using the non-linear features mining and ensemble learning (EL) approach. We first diminish the noise and artifacts from the input EEG-fNIRS signals using digital filtering. After that, we use the signals for non-linear features mining. These features are “Fractal Dimension” (FD), “Higher Order Spectra” (HOS), “Recurrence Quantification Analysis” (RQA) features, and Entropy features. Onward, the Genetic Algorithm (GA) is employed for Features Selection (FS). Lastly, we employ a novel Machine Learning (ML) technique using several algorithms namely, the “Naïve Bayes” (NB), “Support Vector Machine” (SVM), “Random Forest” (RF), and “K-Nearest Neighbor” (KNN). These classifiers are combined as an ensemble for recognizing the intended brain activities. The applicability is tested by using a publicly available multi-subject and multiclass EEG-fNIRS dataset. Our method has reached the highest accuracy, F1-score, and sensitivity of 95.48%, 97.67% and 97.83% respectively.
  • Noise Effects on a Proposed Algorithm for Signal Reconstruction and Bandwidth Optimization

    Ahmed F. Ashour; Ashraf A. M. Khalaf; Hussein, Aziza; Hesham F. A. Hamed; Ashraf Ramadan; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; et al. (Institute of Advanced Engineering and Science (IAES), 2023-06-01)
    The development of wireless technology in recent years has increased the demand for channel resources within a limited spectrum. The system's performance can be improved through bandwidth optimization, as the spectrum is a scarce resource. To reconstruct the signal, given incomplete knowledge about the original signal, signal reconstruction algorithms are needed. In this paper, we propose a new scheme for reducing the effect of adding additive white Gaussian noise (AWGN) using a noise reject filter (NRF) on a previously discussed algorithm for baseband signal transmission and reconstruction that can reconstruct most of the signal’s energy without any need to send most of the signal’s concentrated power like the conventional methods, thus achieving bandwidth optimization. The proposed scheme for noise reduction was tested for a pulse signal and stream of pulses with different rates (2, 4, 6, and 8 Mbps) and showed good reconstruction performance in terms of the normalized mean squared error (NMSE) and achieved an average enhancement of around 48%. The proposed schemes for signal reconstruction and noise reduction can be applied to different applications, such as ultra-wideband (UWB) communications, radio frequency identification (RFID) systems, mobile communication networks, and radar systems.
  • A complexity efficient PAPR reduction scheme for FBMC-based VLC systems

    Radwa A Roshdy; Hussein, Aziza; Mohamed M Mabrook; Mohammed A Salem; External Collaboration; NA; NA; NA; Electrical and Computer Engineering; NA; et al. (Polish Academy of Sciences and Association of Polish Electrical Engineers in cooperation with Military University of Technology, 2023-03-31)
    Visible light communication based on a filter bank multicarrier holds enormous promise for optical wireless communication systems, due to its high-speed and unlicensed spectrum. Moreover, visible light communication techniques greatly impact communication links for small satellites like cube satellites, and pico/nano satellites, in addition to inter-satellite communications between different satellite types in different orbits. However, the transmitted visible signal via the filter bank multicarrier has a high amount of peak-toaverage power ratio, which results in severe distortion for a light emitting diode output. In this work, a scheme for enhancing the peak-to-average power ratio reduction amount is proposed. First, an algorithm based on generating two candidates signals with different peakto-average power ratio is suggested. The signal with the lowest ratio is selected and transmitted. Second, an alternate direct current-biased approach, which is referred to as the addition reversed method, is put forth to transform transmitted signal bipolar values into actual unipolar ones. The performance is assessed through a cumulative distribution function of peak-to-average power ratio, bit error rate, power spectral density, and computational complexity. The simulation results show that, compared to other schemes in literature, the proposed scheme attains a great peak-to-average power ratio reduction and improves the bit the error rate performance with minimum complexity overhead. The proposed approach achieved about 5 dB reduction amount compared to companding technique, 5.5 dB compared to discrete cosine transform precoding, and 8 dB compared to conventional direct current bias of an optical filter bank multicarrier. Thus, the proposed scheme reduces the complexity overhead by 15.7% and 55.55% over discrete cosine transform and companding techniques, respectively

View more