Dr. KUPPUSAMY.P.G.
@jnn.edu.in
DEAN - RESEARCH
J.N.N.INSTITUTE OF ENGINEERING, CHENNAI,TAMILNADU.
RESEARCH, TEACHING, or OTHER INTERESTS
Artificial Intelligence, Computer Engineering, Computer Vision and Pattern Recognition, Computer Networks and Communications
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
T. S. Arulananth, P. G. Kuppusamy, Ramesh Kumar Ayyasamy, Saadat M. Alhashmi, M. Mahalakshmi, K. Vasanth, and P. Chinnasamy
Public Library of Science (PLoS)
Semantic segmentation of cityscapes via deep learning is an essential and game-changing research topic that offers a more nuanced comprehension of urban landscapes. Deep learning techniques tackle urban complexity and diversity, which unlocks a broad range of applications. These include urban planning, transportation management, autonomous driving, and smart city efforts. Through rich context and insights, semantic segmentation helps decision-makers and stakeholders make educated decisions for sustainable and effective urban development. This study investigates an in-depth exploration of cityscape image segmentation using the U-Net deep learning model. The proposed U-Net architecture comprises an encoder and decoder structure. The encoder uses convolutional layers and down sampling to extract hierarchical information from input images. Each down sample step reduces spatial dimensions, and increases feature depth, aiding context acquisition. Batch normalization and dropout layers stabilize models and prevent overfitting during encoding. The decoder reconstructs higher-resolution feature maps using "UpSampling2D" layers. Through extensive experimentation and evaluation of the Cityscapes dataset, this study demonstrates the effectiveness of the U-Net model in achieving state-of-the-art results in image segmentation. The results clearly shown that, the proposed model has high accuracy, mean IOU and mean DICE compared to existing models.
T. S. Arulananth, S. Wilson Prakash, Ramesh Kumar Ayyasamy, V. P. Kavitha, P. G. Kuppusamy, and P. Chinnasamy
Institute of Electrical and Electronics Engineers (IEEE)
There is a substantial worldwide effect, both in terms of disease and death, that is caused by pediatric pneumonia, which is a disorder that affects children under the age of five. Even while Streptococcus pneumoniae is the most prevalent agent responsible for this sickness, it may also be brought on by other bacteria, viruses, or fungi. An efficient approach utilizing deep-learning methods to forecast pediatric pneumonia reliably using chest X-ray images has been developed. The current study presents an updated version of the DenseNet-121 deep-learning model developed for identifying scans of pediatric pneumonia. The batch normalization, maximum pooling, and dropout layers introduced into the standard model were done so to improve its accuracy. The activations of the preceding layers are scaled and normalized using batch normalization, leading to a mean value of zero and a variance of one. This helps decrease internal variability during training, which speeds up the training process, promotes model stability, and improves the model’s overall capacity to generalize. Max pooling is a beneficial technique for reducing the number of model parameters, making the model more computationally effective. Meanwhile, dropout is a preventative measure against overfitting by decreasing the co-dependence of neurons. As a result, the network acquires more durable and adaptive features. Classifying instances of pediatric pneumonia with the help of the proposed model resulted in an exceptional accuracy rate of 97.03%.
Peramandai Govindasamy Kuppusamy, Kotteswaran Rajkumar, Rajagopal Maheswar, Soundarapandian Sheeba Rani, and Iraj Sadegh Amiri
Walter de Gruyter GmbH
Abstract This paper focuses on designing of a 10 Gbit/s-10GHz hybrid Orthogonal frequency division multiplexing (FDM) based Radio over Free Space Optics (Ro-FSO) transmission link using optical single sideband modulation format and its performance has been compared using a different number of transmission beams. The proposed link has been simulated and compared using 1-beam, 2-beam, and 4-beam in the system. We show that by using the 4-beam system in the OFDM-Ro-FSO link, a 3000 m range has been achieved reliably, which a notable improvement when compared to previous work is. Also, we show an improved performance of the system by using an enhanced detection mechanism using a Square root module (SRm) at the receiver side.
Hazem M. El-Hageen, P.G. Kuppusamy, Aadel M. Alatwi, M. Sivaram, Z. Ahamed Yasar, and Ahmed Nabih Zaki Rashed
Walter de Gruyter GmbH
Abstract Different types of laser source modulation techniques have been used in various applications depending on the objective. As optical systems extract the laws and the best solutions from experiments and simulations, the present study uses simulation software with different modulation types so the output signals can be compared. The modulators used are Mach-Zehnder, which is an external modulator, and electro-absorption modulator and laser rate equation modulator, which are direct modulators. All these types have an optical link multimode (MM) fiber with a photodiode in the receiver end that can be modeled. The input and output signals are analyzed using different types of modulations.
S. V. Tresa Sangeetha, P. G. Kuppusamy, V. Chandrasekaran, and A. Sangeerani Devi
AIP Publishing
Vinola. C, Godwin Premi, P. Solainayagi, C. Srinivasan, and P.G. Kuppusamy
IEEE
In this paper, a comprehensive framework for protecting sensitive healthcare information in Internet of Things (IoT)-enabled Wireless Sensor Network (WSN) is presented. Sensitive medical data is protected by the proposed architecture's usage of secure data transmission protocols, encryption, access control, and authentication. It also looks at how the General Data Protection Regulation (GDPR) and other data privacy regulations might affect IoT-enabled healthcare infrastructure. The study results indicate a need for more privacy and security education and implementation throughout the healthcare, technology, government, and user communities. The need of teaching healthcare professionals and patients about the dangers of sharing personal information online and the importance of managing data ethically is discussed. New loopholes and dangers can only be patched if security is regularly assessed, audited, and improved. This research paper sheds light on the problems of privacy and confidentiality in WSN healthcare applications enabled by the Internet of Things. The effort aims to safeguard and defend healthcare IoT adoption and enhance patient care by offering a complete framework that emphasizes regulatory compliance and appropriate data management.
V. Arun, P.G. Kuppusamy, G. Naveen, and P. Santhuja
IEEE
This research study discusses about the prospective future research areas while presenting the state-of-the-art in wireless network cryptography. Elliptic Curve Cryptography (ECC) offers a safe way for interacting nodes to exchange keys. In order to implement ECC, the message is first converted into an affine point on the Elliptic Curve (EC), and then the knapsack method is applied to the ECC-encrypted message across the finite field. This approach presents Knapsack Encryption with ECC Algorithm (KECC) for improving data security in wireless networks. The data is encrypted by using the knapsack algorithm and the ECC algorithm. The necessary encrypted data is forwarded to the receiver via in-between nodes. As a result, it avoids the man-in-the-middle attack. Experimental results prove that the KECC strategy increases the forward ratio. In addition, it reduces the key generation, encryption, as well as decryption computational time.
Eethamakula Kosalendra, K. Krishnamoorthi, S. Diwakaran, P. Vijayakumari, and P.G. Kuppusamy
IEEE
In the situation of Corona virus pandemic, each and every individual suffers a lot, especially the medication team and the respective individuals work as much as harder to safe many people life. Most of the Covid victims are affected by lung oriented and breathing issues. During that period ventilators play a major role to make the people to survive. The medical field requires more and more number of ventilators instantly at the same time to bring back the patient's life from the complicated disease called Corona virus. But practically no hospitals and medical system had such provision to provide thousands of ventilators to patients at the same time. For managing such conditions, a new technology is required to provide sufficient number of devices as the patients required. In this paper, a new robotic ventilator is designed with the help of latest technologies to overcome the situations like Covid-19. This robotic ventilation systems use parts that are widely accessible across the world and that parts are commonly found in commonplace appliances as well as services. This system do not require for any unique production techniques and f or the contemporary pandemic, many solutions have been developed, all with the goal of fulfilling the most fundamental needs for adequate ventilation. But other individuals are opposed using these robotic ventilation systems in real-world circumstances because of their low dependability and failure to satisfy specific medical standards. There are benefits and drawbacks to every implementation of this plan and it's up to designers to work out the kinks. Consequently, by methodical study of the current stock of proposed model, this paper intends to give readers a summary of the main design characteristics that has to be addressed while developing portable ventilation systems. By examining the current research, many parameters are identified that affect efficiency of the device and explained how these aspects must be taken into account for optimal device functioning.
P. G. Kuppusamy, Eethamakula Kosalendra, K. Krishnamoorthi, S. Diwakaran, and P. Vijayakumari
IEEE
An improvement of medical field requires a wide range of support from Artificial Intelligence (AI) system and several learning mechanisms. In such case a logic of Medical Image Processing (MIP) oriented concepts are providing a huge support to such medical fields to analyze complex cases in easier manner like tumor, cancer and so on. The major complication of people now-a-days is a Lung Nodule affection and it produce a drastic effect in human life as well as many of the people are affected severely without identifying this disease in earlier stages. So that in this paper a new logic is introduced called as Novel Deep Learning Algorithm (NDLA), in which it provides a marvelous support to physicians to analyze the lung nodules easily and provide the exact scenario of the affection in detail. The proposed NDLA algorithm works based on processing the Computed Tomography (CT) images, in which the process is undergoing into several stages such as Pre-Processing, Lung Region Segmentation and Classification. The logic of Deep Learning initially trains all the input dataset lung images in detail based on the mentioned processes like preprocessing and so on. The processed images are maintained into the repository for testing the real-world patient records. Finally the input patient CT images of the lung is analyzed according to the processed dataset images and extracts the exact scenario of the disease as well as report the details in clear way to the respective individual or physician with proper accuracy details. A novel dataset acquired from Kaggle is used in this scenario to produce the best outcome in results and the resulting section of this paper provides the details in clear manner. For all the proposed Novel Deep Learning Algorithm provides a better solution to analyze the lung nodule disease in dense manner and provide a strong support to medical field for analyzing the complications in earlier stages to save lives of many people.
S. Diwakaran, P. Vijayakumari, P.G. Kuppusamy, Eethamakula Kosalendra, and K. Krishnamoorthi
IEEE
The proliferation of IIoT devices for control, monitoring, and processing has been spurred by the advent of 5G networks. With biometric-based user identification, IIoT devices may be protected against unwanted access, keeping production data secure. However, most IIoT biometric authentication solutions do not safeguard template data, putting at risk sensitive biometric information kept as models in centralized dataset. Furthermore, conventional biometric verification is hampered by issues with speed, database storage space, and data transfer. In order to solve these problems, we offer a safe E-fingerprint verification solution using 5G networks. The suggested method centers on the creation of a nullifying fingerprint model, which safeguards the real granular details while also guaranteeing the privacy and security of clients and the content of messages sent among devices and the server across the networks. On three public fingerprint dataset, the suggested authentication model achieves comparable performance while minimizing computational costs and providing quick online matching compared to traditional approaches.
P. Vijayakumari, P.G. Kuppusamy, Eethamakula Kosalendra, K. Krishnamoorthi, and S. Diwakaran
IEEE
Rapid improvements in healthcare services and affordable IoT in the past decade have been a big help in dealing with the issue of fewer medical facilities. Unfortunately, some people still choose not to get immunized, thus fear and reluctance remain a part of human existence despite widespread vaccination initiatives. Therefore, it is important to screen this group of potential spreaders as soon as possible since they may become infected and transfer viruses to others. It is in this context that the pharmaceutical sector might benefit from highly developed health monitoring systems. This work has created and tested a multi-node architecture based on Fog computing to perform real-time initial screening and recording of such individuals, therefore addressing the demand and reducing the unpredictability of the scenario. In addition to capturing photographs of the subject's face, the suggested device also recorded the subject's current body temperature and GPS locations. As an added bonus, the suggested system could upload information to a remote server over the internet. To test the viability of the proposed system, a thorough examination of the existing work environment was carried out, including implementation and evaluations. From the results of the statistical analysis, it was seen that the suggested IoT Fog-enabled ecosystem may be put to good use.
K. Krishnamoorthi, S. Diwakaran, P. Vijayakumari, P.G. Kuppusamy, and Eethamakula Kosalendra
IEEE
The emergence of swarm intelligence approaches has resulted in the development of a workable theoretical computational approach to the simulation, modeling, and optimization of complex systems. This research proposes using a Modified sparrow search algorithm (MSSA) to optimize the coverage WSN. There are three areas of the algorithm that have been improved. We first use the LHS approach to produce the initial population. To further improve the method's convergence efficiency, we offer new optimization equations based on an adaptively adjusted sine cosine algorithm and the Lévy flight strategy. Finally, a unique mutation disturbance mechanism is implemented at the conclusion of every iteration to maximize the individuals with low fitness in the natives. Experimental findings from 13 benchmark functions reveal that the proposed improved method has benefits in convergence pace, and strength, as shown by the stability of its average value and the shortness of the time it takes to reach the optimal solution. In this study, we provide a state-of-the-art optimization framework for the WSN coverage issue using swarm intelligence algorithms and then analyze the efficacy of nine methods.
S. G. Hymlin Rose, P. G. Kuppusamy, B. R. Tapas Bapu, and Muruganantham Ponnusamy
Wiley
AbstractIn this study a highly flexible microwave shielding material was fabricated by solution casting method utilizing Nickel and biocarbon particles in PVA matrix and characterized for mechanical, magnetic, and microwave shielding properties. The main aim of this study was to prove the significant role of magnetic particles in electromagnetic interference (EMI) shielding along with conductive particles. The results show that the addition of Ni‐biocarbon hybrid particle increases the shielding properties up to 56.5 dB at 20 GHz. The magnetic permeability increased gradually with the inclusion of Ni particles with a highest magnetization, coercivity, and retentivity of 1250 E−6 emu, −9000 G, and 1100 E−6 emu. Similarly the mechanical results show that adding biocarbon enhances the composite's mechanical properties. A highest tensile strength, tear strength, elongation, and hardness are noted as 38, 168 MPa, 18.4%, and 36 Shore‐D. Comparatively, the hardness and elongation% of composite designations contains 3 and 5 vol% of hybrid particles have increased by 9% and 26%, respectively, in comparison to composite containing only 5 vol% of biocarbon with PVA. Scanning electron microscope fractography indicates biocarbon particles reduce voids and improve adhesion. These flexible EMI shielding composites could be used in telecommunication and other wave transmitting devices in engineering applications.
C. Vanitha, Anandhavelu Sanmugam, A. Yogananth, M. Rajasekar, P.G. Kuppusamy, and G. Devasagayam
Elsevier BV
V Nagaraju, B R Tapas Bapu, P Bhuvaneswari, R Anita, P G Kuppusamy, and S Usha
Springer Science and Business Media LLC
K. Kala, N. Padmasini, B. Suresh Chander Kapali, and P. G. Kuppusamy
Springer Science and Business Media LLC
Taha Junaid, D. Sumathi, A.N. Sasikumar, S. Suthir, J. Manikandan, Rashmita Khilar, P.G. Kuppusamy, and M. Janardhana Raju
Elsevier BV
R. Anitha, B. R. Tapas Bapu, P. G. Kuppusamy, N. Partheeban, and A. N. Sasikumar
Springer Science and Business Media LLC
M. Rajesh Khanna, A. Karthikeyan, P. G. Kuppusamy, and R. R. Prianka
Springer Science and Business Media LLC
I. S. Amiri, P. G. Kuppusamy, Ahmed Nabih Zaki Rashed, P. Jayarajan, M. R. Thiyagupriyadharsan, and P. Yupapin
Walter de Gruyter GmbH
Abstract High-speed single-mode fiber-optic communication systems have been presented based on various hybrid multiplexing schemes. Refractive index step and silica-doped germanium percentage parameters are also preserved during their technological boundaries of attention. It is noticed that the connect design parameters suffer more nonlinearity with the number of connects. Two different propagation techniques have been used to investigate the transmitted data rates as a criterion to enhance system performance. The first technique is soliton propagation, where the control parameters lead to equilibrium between the pulse spreading due to dispersion and the pulse shrinking because of nonlinearity. The second technique is the MTDM technique where the parameters are adjusted to lead to minimum dispersion. Two cases are investigated: no dispersion cancellation and dispersion cancellation. The investigations are conducted over an enormous range of the set of control parameters. Thermal effects are considered through three basic quantities, namely the transmission data rates, the dispersion characteristics, and the spectral losses.
P.J. Sathish Kumar, Ratna Kamala Petla, K. Elangovan, and P.G. Kuppusamy
Wiley
D. Vijendra Babu, V Chandrasekaran, Mohamed Kuresh Safir, and P G Kuppusamy
IEEE
Deepthi Joseph, A. Karthikeyan, P. G. Kuppusamy, and V. Prabhu
Springer Science and Business Media LLC
P G Kuppusamy, R Sureshkumar, S A Yuvaraj, and E Dilliraj
IOP Publishing
Abstract The research study presents a VLSI design of an effective electrocardiogram data encoding lossless data compression scheme to conserve disk system to minimize channel capacity. As the data compression can save disc space, reduce transfer time, and seized this ability by introducing a memory-less architecture when operating in VLSI at a high data rate. There are two components of the ECG classification technique: an adaptive frequency-domain methodology and bandwidth. An accurate and reduced VLSI compressed algorithm design has been introduced. The current VLSI architecture uses a few more procedures to substitute for the various mathematical functions to enhance performance and implemented the VLSI’s architecture to the MIT-BIH atrial fibrillation repository capable of achieving a 2.62 lossless bit compression rate. Also, the VLSI structure uses a gate count of 5.1 K.
R Sureshkumar, S A Yuvaraj, P G Kuppusamy, and E Dilliraj
IOP Publishing
AbstractA significant method to identify and track early cancer of the breast in clinical practise is histopathological image analysis. The diagnosis of breast cancer is still facing issues with an open in healthcare sector, however, with a limited quality. We are creating a classification system based on histological picture pictures, integrating deep learning with mechanical methodologies of learning, in sequence to enhance the prediction of early recognized breast cancer and to minimize the work pressure of physicians. In particular, through pre-trained Deep Convection Neural Networks, we build a multi-network extraction model, create an efficient method of reducing features and train an ecosystem supporting vector machine (E-SVM). Next, we use scale transformation and colour improvement approaches to prepare histological pictures. Second, four pre-trained DCNNs extract the multi-network functionality. Thirdly, the Dual Network Orthogonal Low-Rank Learning (DOLL) role selection approach is further introduced to increase efficiency and to reduce unnecessary efficiencies. An E-SVM is at last instructed by melded usefulness and casting a voting procedure for characterization, what isolates the pictures into four gatherings considerate, in situ carcinomas, obtrusive carcinomas, and ordinary. The suggested procedure is tested by us for the public ICIAR 2018 Challenges Data Set on histology photographs of breast cancer. Our approach can offer very promising productivity and underperform province approaches through analytical outcomes.