Gaurav Singal
@nsut.ac.in
Assistant Professor, Faculty of Inter Disciplinary Studies
Netaji Subhas University of Technology Delhi
Gaurav Singal (SMIEEE) is an Assistant Professor in Computer Science Engineering Department at NSUT Delhi. He obtained his PhD and M. Tech. in the Computer Science Engineering department from Malaviya National Institute of Technology, Jaipur, India. He received research grants from the Department of Science and Technology, Uttar Pradesh on women security and the Department of Biotechnology on Assistive devices. He is actively working in research and teaching for the last 10 years and published a number of reputed conferences and journals. He is a member of the scientific society IEEE and ACM. He is certified as an NVIDIA Deep learning institute ambassador and UIPath RPA Advanced Developer. His research interests include Wireless and Ad hoc Networking, Internet of Things, Edge computing, Applied Deep learning and Reinforcement learning.
EDUCATION
BTech, MTech, PhD
RESEARCH, TEACHING, or OTHER INTERESTS
Computer Engineering, Artificial Intelligence, Computer Networks and Communications
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Abhinav Tomar, Raj Anwit, Piyush Nawnath Raut, and Gaurav Singal
Springer Science and Business Media LLC
Nisha Ahuja, Debajyoti Mukhopadhyay, and Gaurav Singal
Springer Science and Business Media LLC
Shailendra Pratap Singh, Gaurav Dhiman, Sapna Juneja, Wattana Viriyasitavat, Gaurav Singal, Neeraj Kumar, and Prashant Johri
Institute of Electrical and Electronics Engineers (IEEE)
The rapid growth of the Internet of Things (IoT) in the early 21st century has introduced complexities in delivering various services, including Quality-of-Service (QoS) management for smart agriculture sensors. Selecting optimal IoT nodes considering QoS parameters, such as energy consumption, latency, and network coverage area has become challenging. In response, this research proposes an extended form of differential evolution (DE) that incorporates a rapid adaptation approach using optimization-based design. By leveraging dynamic information from IoT devices, the proposed approach enhances exploration and exploitation capabilities, allowing for adaptive adjustment of algorithm parameters and strategies. Additionally, a novel fitness function for energy harvesting in IoT-based applications is introduced. The effectiveness of the proposed algorithm is evaluated in IoT-based applications and an IoT-service framework, with comparative analysis against state-of-the-art algorithms. The results demonstrate that the proposed approach achieves superior performance in energy harvesting QoS, delay, service cost, and maximum coverage area in the IoT-service network. This research contributes to the IoT field by offering an advanced DE algorithm that addresses limitations, providing valuable insights for QoS management in IoT-based services, particularly in the context of smart agriculture sensors.
RETRACTION: V. Baviskar, M. Verma, P. Chatterjee, G. Singal and T. R. Gadekallu, “,” Expert Systems (Early View): , https://doi.org/10.1111/exsy.13359. The above article, published online on 10 June 2023 in Wiley Online Library (wileyonlinelibrary.com), has been retracted by agreement between the journal Editor‐in‐Chief, David Camacho; and John Wiley & Sons Ltd. The article was submitted as part of a guest‐edited special issue. Following publication, it has come to the attention of the journal that parts of the methods in the article lack sufficient detail such that the research cannot be reproduced. A relevant discussion and discrimination for different cardiovascular diseases is missing. The editors have therefore decided to retract this article. The authors disagree with the retraction.
Abhishek Jha, Kartik Goyal, Nalin Semwal, Gaurav Singal, and Abhinav Tomar
Springer Nature Singapore
Riti Kushwaha, Gaurav Singal, and Neeta Nain
Institute of Electrical and Electronics Engineers (IEEE)
Footprint biometrics is one of the emerging techniques, which can be utilized in different security systems. A human footprint has unique traits which is sufficient to recognize any person. Existing work evaluates the shape features and texture features but very few authors have explored minutiae features, hence this article provides a study based on minutiae features. The current State-of-the-art methods utilize machine learning techniques, which suffer from low accuracy in case of poor-quality of data. These machine learning techniques provide approx 97% accuracy while using good quality images but are not able to perform well when we use poor quality images. We have proposed a minutiae matching system based on deep learning techniques which is able to handle samples with adequate noise. We have used Convolution Neural Network for the feature extraction. It uses two different ridge flow estimation methods, i.e., ConvNet-based and dictionary-based. Furthermore, fingerprint-matching metrics are used for footprint feature evaluation. We initially employed a contrastive-based loss function, resulting in an accuracy of 56%. Subsequently, we adapted our approach by implementing a distance-based loss function, which improved the accuracy to 66%.
Vaishali Baviskar, Madhushi Verma, Pradeep Chatterjee, and Gaurav Singal
Elsevier BV
Surbhi Gupta, Gaurav Singal, Deepak Garg, and Sarangapani Jagannathan
Institute of Electrical and Electronics Engineers (IEEE)
Recently introduced deep reinforcement learning (DRL) techniques in discrete-time have resulted in significant advances in online games, robotics, and so on. Inspired from recent developments, we have proposed an approach referred to as Quantile Critic with Spiking Actor and Normalized Ensemble (QC_SANE) for continuous control problems, which uses quantile loss to train critic and a spiking neural network (NN) to train an ensemble of actors. The NN does an internal normalization using a scaled exponential linear unit (SELU) activation function and ensures robustness. The empirical study on multijoint dynamics with contact (MuJoCo)-based environments shows improved training and test results than the state-of-the-art approach: population coded spiking actor network (PopSAN).
Shrey Rastogi, Shivam Sharma, Suraj Kumar, and Gaurav Singal
ACM
The quality of water resources available across the world have been degrading much faster than ever before. Therefore, judicious use of water resources has become an important topic as these resources get evermore scarce. Moreover, efficient treatment and utilization of these resources domestically is another factor to be considered. Our proposed IoT-based system aims to enhance the water utilization in households as in preventing excess use of Reverse Osmosis (RO) and prevent highly contaminated water from entering the house water-storage. The system uses sensors which measure physical characteristics of water – its temperature and Total Dissolved Solids (TDS) which is gathered by the microcontroller and displayed to the user on a smartphone. Based on the value of TDS of water, the system automatically controls the water filter system by directing unfiltered water to the suitable filter (RO or Non-RO) as well as controlling the water-storage tank system.
Kanak Manjari, Madhushi Verma, Gaurav Singal, and Suyel Namasudra
Association for Computing Machinery (ACM)
Scene text detection is complicated and one of the most challenging tasks due to different environmental restrictions, such as illuminations, lighting conditions, tiny and curved texts, and many more. Most of the works on scene text detection have overlooked the primary goal of increasing model accuracy and efficiency, resulting in heavy-weight models that require more processing resources. A novel lightweight model has been developed in this article to improve the accuracy and efficiency of scene text detection. The proposed model relies on ResNet50 and MobileNetV2 as backbones with quantization used to make the resulting model lightweight. During quantization, the precision has been changed from float32 to float16 and int8 for making the model lightweight. In terms of inference time and Floating-Point Operations Per Second, the proposed method outperforms the state-of-the-art techniques by around 30–100 times. Here, well-known datasets, i.e., ICDAR2015 and ICDAR2019, have been utilized for training and testing to validate the performance of the proposed model. Finally, the findings and discussion indicate that the proposed model is more efficient than the existing schemes.
Gaurav Singal, Himanshu Singhal, Riti Kushwaha, Venkataramana Veeramsetty, Tapas Badal, and Sonu Lamba
Springer Science and Business Media LLC
Aryaman Sharma, Harshit Gupta, Tabishi Singh, Gaurav Singal, and Riti Kushwaha
IEEE
In the face of increasing cases of temporary and permanent motion and speech disabilities, the project NayanCom aims to solve the human computer interaction (HCI) problem for people in this demographic. The proposed and implemented approach uses the blinking actions of a user to allow easier communication with their caretakers and medical services for a better patient experience. The project uses computer vision to detect the face and eye blink action of a patient, following which it is processed and fed into the action handler, thus setting up a non-vocal communication medium. Along with the communication aspect, a low cost wireless patient vital monitoring system was also developed wherein the patient’s essential vitals are tracked and reported in real time.
Riti Kushwaha, Himanshu Rawat, Suman Bando, Aman Kumar Dass, Lakshya Panwar, Divya Srivastava, and Gaurav Singal
IEEE
This research paper introduces an innovative solution for enhancing class monitoring by integrating video analytics and advanced deep-learning models for human activity recognition. Through a comparative analysis of two implementation methods, Deep Neural Network (DNN) models emerge as the more efficient and scalable choice, revolutionizing the approach to class monitoring in educational and security contexts. The study evaluates four key deep-learning models, including ConvLSTM, LRCN, and CNN+GRU, highlighting their superior performance over traditional feature-based approaches. Notably, the accuracy metrics for these models are as follows: ConvLSTM achieved an accuracy of 0.62, LRCN exhibited exceptional accuracy at 0.94, and GRU + Convolution Model demonstrated robust accuracy of 0.92. The ultimate goal is the development of an automated monitoring system leveraging deep learning to detect and classify human activities, with significant implications for education and security. These DNN models, known for their cost-effectiveness and adaptability, can be swiftly deployed to monitor classrooms and examination rooms, promptly identifying and reporting suspicious activity. Rigorous performance evaluation using diverse metrics underlines the strengths and weaknesses of these models, marking a significant advancement in automated monitoring systems’ capabilities.
Riya Goyal, Abhinav Tomar, and Gaurav Singal
IEEE
An economic zone is bound to have continuous monitoring and control by an autonomous surveillance system for better production competency and security. Wireless Rechargeable Sensor Networks (WRSNs) have gained popularity to provide reliable and sustainable energy supply for continual network operations. In economic zones, numerous sensor nodes are deployed for diverse activities that require continuous energy supply. However, the integration of WRSNs with UAVs has increased the efficiency of recharging the sensor nodes deployed in areas that are difficult to reach by mobile chargers. Nevertheless, UAVs’ limited battery life limits their performance and range which requires efficient deployment of charging stations. This paper proposes an integration of the Genetic Algorithm (GA) and Ant Colony Optimization (ACO) algorithms to address the charging scheduling problem through efficient deployment of charging stations and finding the optimal charging scheduling path for UAVs. Simulation results demonstrate that the proposed scheme produces a shorter planning path, leading to a longer network lifespan than alternative approaches.
Rhythm Narula, Pulkit Barwal, and Gaurav Singal
IEEE
Many buildings are sensitive to earthquake and hurricanes, and it is easy to generate large deflection and dynamic response even through daily motion of a huge crowd. There are also structures with inappropriate distributions of strength and stiffness which performs poorly in case of any seismic activity and are one of the major causes of collapses leading to loss of lives. With these issues, there is a growing demand for the exploration of different methodologies to detect the prime areas in the structures that require improvement and repair, and with these early detection techniques, we can take suitable action that can, in turn, save a lot of lives. Previous research in this field mainly revolved around checking concrete strength and rust which was insufficient. Other devices which exist are expensive and they has to be equipped while constructing the building. In this paper we have summarised our study on finding the structural strength of buildings with the help of MPU6050 gyroscope. The MPU6050 gyroscope is a 3-axis gyroscope and a 3-axis accelerometer. The sensor measures the angular rotation from its axis and transfers it to the smartphone using a Bluetooth module.
Kanak Manjari, Animesh, Madhushi Verma, and Gaurav Singal
IEEE
The computer vision technique known as object detection is used to determine the location of an object in an image or video. Recent advances in a number of areas, including computer vision, speech recognition, and natural language processing have been made using deep learning approaches. Convolution neural networks (CNNs), which have several layers, are used to detect objects. A deep learning model that uses CNN to identify various surface regions has been suggested. The study of tracks, walkways, playgrounds, highways, and muddy areas is a pertinent and significant aspect of object detection for Visually Impaired (VI) individuals. In addition to adding indirect vision to the blind individual, this model recognizes surface regions and alerts them using text-to-sound conversion API. In this study, transfer learning was carried out utilizing Faster RCNN (Region Based Convolutional Neural Networks) with Inception V2 and ResNet50 model weights. The accuracy of both trained models was the same, while Faster RCNN Inception V2 had the lowest loss (0.02).
Riti Kushwaha, Gaurav Singal, and Neeta Nain
IEEE
A human footprint has all the properties like fingerprint which are required for authentication. It has all the desirable properties including uniqueness, universality, and invariability, yet not very widely used for authentication purpose. Hence exploring this novel biometric system is very important. The article explores the singularity available in a footprint image, it is further utilized while performing footprint classification. The main application of footprint classification is to reduce the sample size while performing the authentication. Four different singularities are manually marked on the collected dataset, furthermore, it is trained using different CNN models VGGNet16 and ResNet. A VGGNet-16 model has 16 convolution layer of uniform architecture,with 138 million parameters per layer, while a ResNet50 has 152 layers which is 8×8 times deeper than VGGNet-16. These models are tested on poor-quality dataset (most of the time we are not able to collect clear footprint images), it provides the accuracy of 80%.
Divyanshu Rao, Aryan Sharma, Hrishikesh Shah, Gaurav Singal, Riti Kushwaha, and Vandana Bhatia
IEEE
The primary vision behind this paper is to develop a one-stop solution from the consumer end to the supplier of water in cities. The solution is being enabled through a kit that can be installed in the water tanks. Besides this, the consumer and the water supplying authorities can monitor the functionalities and reports remotely. The kit is built using various sensors which measure the most crucial parameters to determine the quality of water that is being loaded onto the tank. The sensors used here are powerful to measure Turbidity, pH, Temperature, and quantity of the water. All these powerful sensors have been interfaced with the microcontroller unit which can transmit the data remotely.
Kanak Manjari, Madhushi Verma, Gaurav Singal, and Vinay Chamola
Informa UK Limited
Vaishali Baviskar, Madhushi Verma, Pradeep Chatterjee, Gaurav Singal, and Thippa Reddy Gadekallu
Wiley
AbstractRecently, machine learning methods have been successfully used for the prediction of cardiovascular disease. Early diagnosis and prediction is necessary for giving effective treatment to avoid higher mortality rates. Several classification algorithms have been developed recently which satisfy the need, but show limited accuracy while predicting the heart disease. Hence, the focus of this study is on early prediction of heart disease and to improve the accuracy of prediction using benchmark heart disease datasets such as UCI Cleveland dataset and Heart disease clinical dataset by implementing effective classification and optimization algorithms. Optimization algorithms generally exhibit the benefit of dealing with complex non‐linear issues with better adaptability and flexibility. The Emperor penguin optimization algorithm, which can select the best features for classification has been utilized in this study to improve the efficiency, minimize reconstruction errors, and increase the quality of heart disease classification. Further, the newly developed stacked sparse convolutional neural network based auto encoder (SSC‐AE) classification algorithm has been employed for significant feature classification with higher robustness and efficacy. Accuracy, Area Under Curve (AUC), and F1 score are some of the measures used to compare the outcomes of several machine learning algorithms to those of the proposed model in this study. The results show that the proposed model, SSC‐AE, is superior to other classification models.
Pranav Chaudhary, Vishal Singh, Aayush Karjee, Gaurav Singal, and Abhinav Tomar
Springer Nature Switzerland
Aakash Katiyar, Akash Singh, Mohammad Parvez, Nitin Rajora, Lokesh K Sharma, and Gaurav Singal
IEEE
Home automation is a technology that is often viewed as a luxury and not a necessity. The means of achieving home automation is always considered expensive and inconvenient in the long run. Due to this there is no success in achieving this technology. Our aim is to make life much easier, more comfortable, and much more convenient and reduce our daily workload. To achieve this, our approach was to make a highly compact and simplified product that would be affordable to middle-class families by using familiar applications like blynk and machine learning - for making the home automation personalized to every specific user. Not only that but also to take utmost care of the privacy of our users.
Saon Banerjee, Gaurav Singal, Sarathi Saha, Himanshu Mittal, Manu Srivastava, Asis Mukherjee, Sayak Mahato, Barnali Saikia, Sudipta Thakur, Suman Samanta,et al.
Springer Science and Business Media LLC
Kaushik Ray, Vipin Pal, Gaurav Singal, and Soumen Moulik
Elsevier BV
Surbhi Gupta, Gaurav Singal, Deepak Garg, and Swagatam Das
Institute of Electrical and Electronics Engineers (IEEE)
Artificial agents are used in autonomous systems such as autonomous vehicles, autonomous robotics, and autonomous drones to make predictions based on data generated by fusing the values from many sources such as different sensors. Malfunctioning of sensors was noticed in the robotics domain. The correct observation from sensors corresponds to the true estimate of the dimension value of the state vector in deep reinforcement learning (DRL). Hence, noisy estimates from these sensors lead to dimensionality impairment in the state. DRL policies have shown to stagger its decision by the wrong choice of action in case of adversarial attack or modeling error. Hence, it is necessary to examine the effect of dimensionality perturbation on neural policy. In this regard, we analyze whether subtle dimensionality perturbation that occurs due to the noise in the source of input at the testing time distracts agent decisions. Also, we propose an RSAC (robust soft actor-critic) approach that uses a noisy state for prediction, and estimates target from nominal observation. We find that the injection of such noisy input during training will not hamper learning. We have done our simulation in the OpenAI gym MuJoCo (Walker2d-V2) environment and our empirical results demonstrate that the proposed approach competes for SAC’s performance and makes it robust to test time dimensionality perturbation.