Summary
Dr. Muhammad Farrukh Qureshi is an Assistant Professor in the Department of Electrical Engineering at Namal University, Mianwali, bringing over nine years of experience in the field. He earned his BS in Electronic Engineering from IIUI and MS in Electrical Engineering from Riphah International University, specializing in Signal Processing. Dr. Farrukh completed his PhD in Electrical Engineering from Riphah International University in 2024, focusing on Deep Learning in EMG signals. He has published 19 research papers across various domains, including AI-based healthcare, signal processing, cancer cell classification, EEG, EMG, etc. His work has garnered over 210 citations and an h-index of 9. Dr. Farrukh's current research interests include innovative machine learning applications in healthcare, agriculture, energy sectors, federated learning, and developing AI solutions tailored to rural environments.
Academic Background
PhD in Electrical Engineering
(Classification of Physiological Signals for Human Computer Interaction using Deep Learning)
Riphah International University, Islamabad.
2024
MS in Electrical Engineering
(Time-Frequency Signal Processing of Self-Mixing Laser Sensor for Vibration Measurement)
Riphah International University, Islamabad.
2014
BS in Electronics Engineering
(Automatic Secured Storage Machine)
International Islamic University, Islamabad.
2012
Experience
Assistant Professor
Namal University Mianwali
09-Sep-2024 - continue
Senior Lecturer
Riphah International University, Islamabad.
01-Sep-2022 - 06-Sep-2024
Lecturer
Riphah International University, Islamabad.
08-May-2015 - 31-Aug-2022
Technical Executive
Wi-tribe Pakistan (Pvt.) Ltd.
29-Nov-2014 - 07-May-2015
Electrical Engineer
Tesla Industries (Pvt.) Ltd.
01-Apr-2014 - 01-Jul-2014
Intern (National Internship Program)
Ministry of IT & Telecommunications, Islamabad
01-Jan-2013 - 31-Dec-2013
Intern (Production)
Pakistan Accumulators (Pvt) Ltd.
01-Oct-2012 - 31-Dec-2012
Journal Publications
EEG electrode setup optimization using feature extraction techniques for neonatal sleep state classification
31-Jan-2025
An optimal arrangement of electrodes during data collection is essential for gaining a deeper understanding of neonatal sleep and assessing cognitive health in order to reduce technical complexity and reduce skin irritation risks. Using electroencephalography (EEG) data, a long-short-term memory (LSTM) classifier categorizes neonatal sleep states. An 16,803 30-second segment was collected from 64 infants between 36 and 43 weeks of age at Fudan University Children's Hospital to train and test the proposed model. To enhance the performance of an LSTM-based classification model, 94 linear and nonlinear features in the time and frequency domains with three novel features (Detrended Fluctuation Analysis (DFA), Lyapunov exponent, and multiscale fluctuation entropy) are extracted. An imbalance between classes is solved using the SMOTE technique. In addition, the most significant features are identified and prioritized using principal component analysis (PCA). In comparison to other single channels, the C3 channel has an accuracy value of 80.75% ± 0.82%, with a kappa value of 0.76. Classification accuracy for four left-side electrodes is higher (82.71% ± 0.88%) than for four right-side electrodes (81.14% ± 0.77%), while kappa values are respectively 0.78 and 0.76. Study results suggest that specific EEG channels play an important role in determining sleep stage classification, as well as suggesting optimal electrode configuration. Moreover, this research can be used to improve neonatal care by monitoring sleep, which can allow early detection of sleep disorders. As a result, this study captures information effectively using a single channel, reducing computing load and maintaining performance at the same time. With the incorporation of time and frequency-domain linear and nonlinear features into sleep staging, newborn sleep dynamics and irregularities can be better understood.
Automated lumpy skin grading in bovine images using novel deep convolutional neural networks
05-Nov-2024
The diagnosis and prevention of lumpy skin disease, a viral ailment affecting cattle and buffalo, present significant financial implications for the livestock industry. Traditional methods for identifying lumpy skin disease rely on manual visual inspection by veterinarians, which can be labor-intensive, subjective, and prone to errors. To address these challenges, this study proposes a novel deep convolutional neural network (DCNN) model for the automatic recognition and grading of lumpy skin disease from bovine images. The primary contributions of this research include the development of a DCNN architecture specifically tailored for this task, comprising five convolutional layers, five max pooling layers, two fully connected layers with ReLU activation, and a final fully connected layer with softmax activation. The model’s detection accuracy is further enhanced by applying image cropping and patching techniques, which divide each input image into 12 patches to improve local feature extraction. The proposed model was trained and tested using a publicly available dataset from Kaggle. Comparative analysis was conducted against several state-of-the-art models, including InceptionV3, ResNet50, MobileNetV3, VGG19, and Xception. The DCNN model demonstrated superior performance, achieving the highest validation accuracy of 0.96875, outperforming the compared models in terms of accuracy, precision, recall, and F1 score. Additionally, the study explores the potential of transitioning from binary to multiclass classification, which would allow for the assessment of the severity of lumpy skin disease. This future direction aims to provide more nuanced and actionable information for veterinary diagnostics. The significance of this research lies in its potential to offer an objective, efficient, and scalable solution for early disease detection and prevention in livestock, thereby presenting considerable economic benefits for farmers and the livestock industry as a whole. The methodology, including data preprocessing, augmentation, model training, and evaluation, is comprehensively detailed to ensure reproducibility and to highlight the robustness of the proposed approach.
Augmenting Aquaculture Efficiency through Involutional Neural Networks and Self-Attention for Oplegnathus Punctatus Feeding Intensity Classification from Log Mel Spectrograms
05-Jun-2024
Understanding the feeding dynamics of aquatic animals is crucial for aquaculture optimization and ecosystem management. This paper proposes a novel framework for analyzing fish feeding behavior based on a fusion of spectrogram-extracted features and deep learning architecture. Raw audio waveforms are first transformed into Log Mel Spectrograms, and a fusion of features such as the Discrete Wavelet Transform, the Gabor filter, the Local Binary Pattern, and the Laplacian High Pass Filter, followed by a well-adapted deep model, is proposed to capture crucial spectral and spectral information that can help distinguish between the various forms of fish feeding behavior. The Involutional Neural Network (INN)-based deep learning model is used for classification, achieving an accuracy of up to 97% across various temporal segments. The proposed methodology is shown to be effective in accurately classifying the feeding intensities of Oplegnathus punctatus, enabling insights pertinent to aquaculture enhancement and ecosystem management. Future work may include additional feature extraction modalities and multi-modal data integration to further our understanding and contribute towards the sustainable management of marine resources.
EMG gesture signal analysis towards diagnosis of upper limb using dual-pathway convolutional neural network
24-Apr-2024
This research introduces a novel dual-pathway convolutional neural network (DP-CNN) architecture tailored for robust performance in Log-Mel spectrogram image analysis derived from raw multichannel electromyography signals. The primary objective is to assess the effectiveness of the proposed DP-CNN architecture across three datasets (NinaPro DB1, DB2, and DB3), encompassing both able-bodied and amputee subjects. Performance metrics, including accuracy, precision, recall, and F1-score, are employed for comprehensive evaluation. The DP-CNN demonstrates notable mean accuracies of 94.93 ± 1.71% and 94.00 ± 3.65% on NinaPro DB1 and DB2 for healthy subjects, respectively. Additionally, it achieves a robust mean classification accuracy of 85.36 ± 0.82% on amputee subjects in DB3, affirming its efficacy. Comparative analysis with previous methodologies on the same datasets reveals substantial improvements of 28.33%, 26.92%, and 39.09% over the baseline for DB1, DB2, and DB3, respectively. The DP-CNN’s superior performance extends to comparisons with transfer learning models for image classification, reaffirming its efficacy. Across diverse datasets involving both able-bodied and amputee subjects, the DP-CNN exhibits enhanced capabilities, holding promise for advancing myoelectric control.
Improving Thyroid Disorder Diagnosis via Ensemble Stacking and Bidirectional Feature Selection.
01-Mar-2024
Thyroid disorders represent a significant global health challenge with hypothyroidism and hyperthyroidism as two common conditions arising from dysfunction in the thyroid gland. Accurate and timely diagnosis of these disorders is crucial for effective treatment and patient care. This research introduces a comprehensive approach to improve the accuracy of thyroid disorder diagnosis through the integration of ensemble stacking and advanced feature selection techniques. Sequential forward feature selection, sequential backward feature elimination, and bidirectional feature elimination are investigated in this study. In ensemble learning, random forest, adaptive boosting, and bagging classifiers are employed. The effectiveness of these techniques is evaluated using two different datasets obtained from the University of California Irvine-Machine Learning Repository, both of which undergo preprocessing steps, including outlier removal, addressing missing data, data cleansing, and feature reduction. Extensive experimentation demonstrates the remarkable success of proposed ensemble stacking and bidirectional feature elimination achieving 100% and 99.86% accuracy in identifying hyperthyroidism and hypothyroidism, respectively. Beyond enhancing detection accuracy, the ensemble stacking model also demonstrated a streamlined computational complexity which is pivotal for practical medical applications. It significantly outperformed existing studies with similar objectives underscoring the viability and effectiveness of the proposed scheme. This research offers an innovative perspective and sets the platform for improved thyroid disorder diagnosis with broader implications for healthcare and patient well-being.
Efficient thyroid disorder identification with weighted voting ensemble of super learners by using adaptive synthetic sampling technique
14-Aug-2023
There are millions of people suffering from thyroid disease all over the world. For thyroid cancer to be effectively treated and managed, a correct diagnosis is necessary. In this article, we suggest an innovative approach for diagnosing thyroid disease that combines an adaptive synthetic sampling method with weighted average voting (WAV) ensemble of two distinct super learners (SLs). Resampling techniques are used in the suggested methodology to correct the class imbalance in the datasets and a group of two SLs made up of various base estimators and meta-estimators is used to increase the accuracy of thyroid cancer identification. To assess the effectiveness of our suggested methodology, we used two publicly accessible datasets: the KEEL thyroid illness (Dataset1) and the hypothyroid dataset (Dataset2) from the UCI repository. The findings of using the adaptive synthetic (ADASYN) sampling technique in both datasets revealed considerable gains in accuracy, precision, recall and F1-score. The WAV ensemble of the two distinct SLs that were deployed exhibited improved performance when compared to prior existing studies on identical datasets and produced higher prediction accuracy than any individual model alone. The suggested methodology has the potential to increase the accuracy of thyroid cancer categorization and could assist with patient diagnosis and treatment. The WAV ensemble strategy computational complexity and the ideal choice of base estimators in SLs continue to be constraints of this study that call for further investigation.
Efficient energy management for household: Optimization-based integration of distributed energy resources in smart grid
07-Aug-2023
Energy demand is increasing globally due to the growing human population and progressive lifestyle. The adequate use of available energy resources, including renewable, contributes to a country’s economic sustainability and future development. Optimization-based energy management and cost minimization plays a significant role in overcoming energy crises in less developed countries. In this paper, an optimization–based dynamic energy management technique for smart grids is developed based on the integration of available renewable resources and variable consumer demand, distinctive to underdeveloped countries. Consumer demand is classified into fixed, flexible, and highly variable based on population characteristics. In this work, we developed a Dynamic Multiple Knapsack DMKNS algorithm, which automatically schedules energy provision to various users by optimally accounting for the available resources (Grid and Renewable). The proposed method provides a low-cost solution by maintaining a constant energy supply while preserving consumer comfort and grid stability. The simulation results with various intermittent availability of resources using MKNS show a saving up to 50% for a variable energy demand user. The proposed method is general and can also be applied to various underdeveloped regions with similar consumer demand and statistics.
Automated uterine fibroids detection in ultrasound images using deep convolutional neural networks
20-May-2023
Fibroids of the uterus are a common benign tumor affecting women of childbearing age. Uterine fibroids (UF) can be effectively treated with earlier identification and diagnosis. Its automated diagnosis from medical images is an area where deep learning (DL)-based algorithms have demonstrated promising results. In this research, we evaluated state-of-the-art DL architectures VGG16, ResNet50, InceptionV3, and our proposed innovative dual-path deep convolutional neural network (DPCNN) architecture for UF detection tasks. Using preprocessing methods including scaling, normalization, and data augmentation, an ultrasound image dataset from Kaggle is prepared for use. After the images are used to train and validate the DL models, the model performance is evaluated using different measures. When compared to existing DL models, our suggested DPCNN architecture achieved the highest accuracy of 99.8 percent. Findings show that pre-trained deep-learning model performance for UF diagnosis from medical images may significantly improve with the application of fine-tuning strategies. In particular, the InceptionV3 model achieved 90% accuracy, with the ResNet50 model achieving 89% accuracy. It should be noted that the VGG16 model was found to have a lower accuracy level of 85%. Our findings show that DL-based methods can be effectively utilized to facilitate automated UF detection from medical images. Further research in this area holds great potential and could lead to the creation of cutting-edge computer-aided diagnosis systems. To further advance the state-of-the-art in medical imaging analysis, the DL community is invited to investigate these lines of research. Although our proposed innovative DPCNN architecture performed best, fine-tuned versions of pre-trained models like InceptionV3 and ResNet50 also delivered strong results. This work lays the foundation for future studies and has the potential to enhance the precision and suitability with which UF is detected.
Enhancing ductal carcinoma classification using transfer learning with 3D U-net models in breast cancer imaging
27-Mar-2023
Breast cancer ranks among the leading causes of death for women globally, making it imperative to swiftly and precisely detect the condition to ensure timely treatment and enhanced chances of recovery. This study focuses on transfer learning with 3D U-Net models to classify ductal carcinoma, the most frequent subtype of breast cancer, in histopathology imaging. In this research work, a dataset of 162 microscopic images of breast cancer specimens is utilized for breast histopathology analysis. Preprocessing the original image data includes shrinking the images, standardizing the intensities, and extracting patches of size 50 × 50 pixels. The retrieved patches were employed to construct a basic 3D U-Net model and a refined 3D U-Net model that had been previously trained on an extensive medical image segmentation dataset. The findings revealed that the fine-tuned 3D U-Net model (97%) outperformed the simple 3D U-Net model (87%) in identifying ductal cancer in breast histopathology imaging. The fine-tuned model exhibited a smaller loss (0.003) on the testing data (0.041) in comparison to the simple model. The disparity in the training and testing accuracy reveals that the fine-tuned model may have overfitted to the training data indicating that there is room for improvement. To progress in computer-aided diagnosis, the research study also adopted various data augmentation methodologies. The experimental approach that was put forward achieved state-of-the-art performance, surpassing the benchmark techniques used in previous studies in the same field, and exhibiting greater accuracy. The presented scheme has promising potential for better cancer detection and diagnosis in practical applications of mammography.
E2CNN: An efficient concatenated cnn for classification of surface emg extracted from upper limb
15-Mar-2023
Surface electromyography is a bioelectrical indicator that emerges during muscle contraction and has been widely used in a variety of clinical applications. Several prosthetic control applications can benefit from the analysis based on the classification of surface electromyography (sEMG) signals. However, for the real-time application of upper limb prosthesis, the EMG-based systems need robust performance and rapid response behavior. In this study, we propose an efficient concatenated convolutional neural network (E2CNN) for classification of sEMG extracted from the upper limb. We have tested and validated the performance of the proposed E2CNN on two datasets: a longitudinal dataset comprising ten nondisabled (healthy) subjects and six transradial amputee subjects and spanning the data collected for a period of seven days; and the publicly available NinaPro DB1 dataset. The raw sEMG signals are converted into Log-Mel spectrograms (LMSs). This model combines the input layers with the output of each convolutional block using concatenation layers. The proposed E2CNN when applied to LMS-based images provides a good response time with high-performance accuracy of 98.31% ± 0.5% and 97.97% ± 1.41% for both nondisabled and amputee subjects. When applied to NinaPro DB1, the proposed E2CNN has attained a mean accuracy of 91.27%, an increase by 24.67% with respect to the baseline CNN model. The results show that the achieved results are comparable to those obtained using stacked sparse autoencoders (SSAEs) and other CNN models; however, E2CNN is associated with reduced training and prediction times, making it a potential candidate for real-time classification of sEMG based on LM spectrogram images.
Effective kernel?principal component analysis based approach for wisconsin breast cancer diagnosis
14-Jan-2023
This work aims to identify cancerous (malignant) and non-malignant (non-cancerous) cells in a breast cancer database. Wisconsin breast cancer data (WBC) was utilized and obtained from the University of California, Irvine's machine learning repository. The proposed approach involves the Naive bayes algorithm with Gaussian distribution of the function in combination with Chi-squared-based attributes selection approach. This experimentation has been done after reducing the dimensional space of the used data with extended Kernel Principal Component Analysis (K-PCA). Five different kernels in K-PCA have been tested after the implementation of necessary pre-processing techniques. The performance assessment of the proposed system has been evaluated based on confusion matrix-based accuracy, precision, sensitivity, and specificity. Our proposed methodology with six selected feature and sigmoid K-PCA attained the best accuracy of 99.28%. This result outer performs many state-of-the-art studies recently published on the identical dataset.
Spectral image-based multiday surface electromyography classification of hand motions using CNN for human–computer interaction
20-Sep-2022
Physiological signals such as electromyography (EMG) have been used in human–computer interaction (HCI) for medical applications. Wearable prostheses, such as robotic limbs, have seen a surge in popularity because of technological advancements in myoelectric interfaces. In spite of encouraging achievements with pattern-recognition-based control systems, user acceptability of prosthetic hands still needs improvement in control robustness. The purpose of this research is to compare multiday surface EMG (sEMG) recordings and measure the performance of convolutional neural network (CNN) to enhance myoelectric control. The performance metrics used in this study are accuracy, macro weighted precision (MWP), macro weighted recall (MWR), and macro F1 -score for eight able-bodied (healthy and nondisabled) and four amputee subjects. Using Mel-spectrogram-based sEMG data from both the able-bodied (healthy and nondisabled) and amputee participants, our proposed CNN has achieved a mean classification accuracy of 99.42% ± 0.42% and 98.00% ± 0.58% for the able-bodied (healthy and nondisabled) and amputee subjects for the within-day analysis, respectively. The proposed CNN outperformed other classifiers ( p≤0.05 ) in the between-day analysis for twofold (65.88% ± 10.1% and 58.37% ± 9.11%) and for sevenfold validation (88.73% ± 1.43% and 77.35% ± 2.72%) using sEMG recordings from the able-bodied (healthy and nondisabled) and amputee subjects, respectively. The proposed CNN is compared with the pretrained transfer learning (TL) models and has achieved higher accuracy with lower computational cost. The results demonstrate that CNN can considerably increase the effectiveness of the pattern recognition myoelectric control schemes and can extract deep information from EMG data.
Spectral Processing of Self-Mixing Interferometric Signal Phase for Improved Vibration Sensing under
Weak-and Moderate-Feedback Regime
13-Aug-2019
In this paper, spectral processing of laser Self-Mixing (SM) interferometric signal phase has been carried out allowing better measurement accuracy for harmonic and arbitrarily shaped vibrations for an optical feedback based SM interferometric Laser Diode (LD) sensor. The resulting algorithm not only improves the measurement accuracy but also reduces the processing time (by a factor 3.45) as compared with a previous time-domain based displacement retrieval technique called the Phase Unwrapping Method (PUM). Fourier series based analysis of laser feedback phase is carried out to determine processing limits. The proposed algorithm has also been found to be robust against variation in optical feedback coupling C as well as additive noise. This use of spectral analysis not only increases the measurement accuracy but also retrieves information about target movement harmonics which can also be used in modal analysis applications of remote mechanical targets. Using an SM vibration sensor based on a LD emitting at 785 nm, this technique has provided an average RMS error of 12.5 nm (while that of PUM is 39 nm RMS) for harmonic target vibrations of 5 μm amplitude. For reduced range of 1 <; C <; 2, an average RMS error of ~8 nm (~λ/100) is obtained.
Conference Publications
Discrete Wavelet Transform for Laser Diode Self-Mixing Interferometry Processing for Vibrations Measurement
27-Nov-2024
Optical feedback interferometry, known for its efficient resource usage, contrasts traditional two-beam interferometry. Laser diode self-mixing interferometry (SM) is a prevalent remote sensing approach. Despite providing insights into target and laser diode dynamics, noise compromises SM interferometry’s efficacy. Leveraging time-frequency signal processing reduces noise in self-mixing signals. This paper presents a comprehensive investigation into the processing of self-mixing laser sensor signals, specifically focusing on leveraging the discrete wavelet transform (DWT) for enhanced analysis. A Wavelet Transform-based Phase Unwrapping Technique is introduced, surpassing traditional methods like the Fast Fourier Transform, especially in handling non-stationary self-mixing (SM) signals. The technique’s key strength lies in its ability to simultaneously retain frequency and temporal information, effectively addressing challenges associated with dynamic signals. The proposed method ensures precise outcomes for stationary and non-stationary signals, marking a significant advancement in laser sensor signal processing. By harnessing DWT’s capabilities, the approach offers heightened accuracy and versatility, shaping the trajectory of SM signal analysis. Incorporating the Discrete Wavelet Transform (DWT) for SMI, our study achieved remarkable accuracy improvements. DWT-based processing resulted in an error of 12.77nm as compared to 60.98nm for RPUM, 5.95nm for FFT for stationary signals, and 11.98nm as compared to 64.64nm for RPUM and 1228nm for FFT for non-stationary signals, marking a significant advancement in laser sensor signal processing. Through this work, we provide a transformative avenue for advancing laser sensor research and elevating signal processing methodologies across disciplines.
Real-Time Weed Segmentation in Tobacco Crops Utilizing Deep Learning on a Jetson Nano
23-May-2024
This paper explores the deployment of a deep learning model on an edge device, such as the Jetson Nano for real-time weed detection in agricultural environments. The U-Net-MobileNetV2 architecture is utilized to achieve accurate and efficient detection of weeds while minimizing computational costs and inference time. The effectiveness of the model in accurately segmenting weed in aerial images of tobacco fields has been demonstrated through a series of simulations and real-time field validations. The methodology proposed in this study demonstrated 96% accuracy, accompanied by a mean intersection over union value of 0.851. Furthermore, the trained model is implemented on the Jetson Nano platform, and real-time validation is showcased. The successful deployment of the model on a mobile setup in the tobacco fields of Mansehra, Pakistan, underscores its practical applicability and relevance to precision agriculture practices.
Deep learning-based classification of wheat leaf diseases for edge devices
27-Nov-2023
This paper proposes a deep learning-based approach for classifying wheat leaf diseases such as stripe rust and septoria for edge devices. The study used a 407 wheat leaf images dataset with three classes: healthy, stripe rust, and septoria. Data augmentation techniques created more training images once the dataset was divided into training, validation, and testing sets. The classification was done using a convolutional neural network (CNN) with a test set accuracy of 98.77%. The outcomes show that deep learning techniques are effective for accurately classifying wheat leaf diseases using cutting-edge devices, with potential for early detection in the field. Future work can include exploring advanced deep-learning techniques and larger datasets to enhance the model’s performance significantly. The proposed method can be used in various applications, such as mobile phone apps, to quickly and accurately detect wheat leaf diseases in the field. The potential impact of the proposed approach is significant, as it can help prevent crop loss and increase crop yield, leading to a more sustainable and food-secure future.
Design and Development of a Soft Robotic Gripper for Precision Control for Biomedical Applications
23-May-2023
Soft robotics offers a captivating frontier within medical engineering, with profound implications spanning various sectors, particularly biomedicine. Our research seeks to explore the intricacies of designing, fabricating, and applying soft robotic grippers in biomedical contexts and applications. Crafted from pliable elastomers and fibers, our pneumatic, soft robotic gripper possesses the unique ability to adapt to the shapes of objects, enabling secure and gentle grasping. The gripper works well because it has pneumatic channels and chambers built in. It also has advanced sensing and control mechanisms that make it work even better, allowing it to interact precisely with biological specimens and medical equipment. Its flexibility and adaptability enable the meticulous and safe manipulation of delicate biological structures, paving the way for groundbreaking advancements in medical procedures. Moreover, by incorporating soft robotic grippers into biomedical research, scientists can conduct finer, controlled manipulations such as tissue sampling, cell manipulation, and targeted drug delivery. This technology dawns a new era of precision and efficacy in biomedical interventions and experimentation, promising transformative breakthroughs in the field.
Leveraging training strategies of artificial neural network for classification of multiday electromyography signals
02-Dec-2022
It is essential to have an improved classification accuracy of target hand movements for the electronic prosthesis in order to work efficiently. As a result, twelve different artificial neural networks (ANN) training strategies have been analyzed, and their performances have been compared to discover the optimal training approach for Electromyography (EMG) signals. The proposed framework was also tested on multiday EMG data to assess its scalability performance. A Wearable MYO wristband is used to collect EMG data from eight participants. The experimental results demonstrate that resilient backpropagation can achieve a classification accuracy of 95%; however, it takes 24 seconds to execute and has a hidden layer size (HLS) of 16. Scaled conjugate gradient, on the other hand, obtained 87% classification accuracy with a 3-second execution time and an HLS of 8.
Estimation of Real-Time Wheeled Mobile Robot (Differential Drive) Motion & Pose with Obstacle Avoidance
16-Aug-2022
This study involves four various steps for the estimation of wheeled mobile robot pose and motion. Our approach involves implementation of specific poses on RVIZ (Visual Simulator). Later step includes the involvement of one real-time robot (TurtleBot3) is used in a differential drive manner. The controlled movement of this robot has been visualized in four different steps. The first phase of the experiment is feed forward control, secondly circular motion control whereas third is a closed loop advancement and final one is a specific pose motion of the robot. Platform used for this task is python with a TurtleBot mobile robot with its navigation software. Experiment first done on the simulation software. Later this experiment has been done in a real time environment. For final pose motion of a robot simulation result is compared with actual real time experiment of TurtleBot. Results shows maximum efficiency and accuracy of our real time TurtleBot.
SVM-based real-time classification of prosthetic fingers using myo armband-acquired electromyography data
25-Oct-2021
In this work we applied real-time classification of prosthetic fingers movements using surface electromyography (sEMG) data. We employed support vector machine (SVM) for classification of fingers movements. SVM has some benefits over other classification techniques e.g. 1) it avoids overfitting, 2)
handles nonlinear data efficiently and 3) it is stable. SVM is employed on Raspberry pi which is a low-cost, credit-card sized computer with high processing power. Moreover, it supports Python which makes it easy to build projects and it has multiple interfaces available. In this paper, our aim is to perform
classification of prosthetic hand relative to human fingers. To assess the performance of our framework we tested it on ten healthy subjects. Our framework was able to achieve mean classification accuracy of 78%.
Android based Internet accessible infant incubator
16-Nov-2019
An infant incubator is meant for the neonatal who
are born premature due to medical complications. This paper
aims to design an android based infant incubator which could
be accessed and controlled via the android application by the
health professional over the Internet. The system designed is
able to collect the information of incubators environment and
store it on an online server. The system could play a significant
role in reducing the mortality rate of the premature infants in
remote areas.
Courses
- Applications of Information and Communication Technologies.
- Electric Machines
- Introduction to Machine Learning
Unmanned Aerial Vehicle for Pesticide Spraying
Won funding of PKR 70,000 from National Grassroot ICT Research Initiative, Ignite-National Technology Fund.
Development of AI model for early diagnosis of Cardiopulmonary
diseases
Awarded funding of PKR 100,000 under FYDP Financing by the Pakistan Engineering Council.
Design and Fabrication of Electric Bike
Won funding of Rs. 70,000 under National Grassroot ICT Research Initiative, Ignite-National Technology Fund.
Smart Fuel Monitoring and Fraud Prevention System for Truck Fleets
Patent submission is in progress.