International Conference on Innovative Computing and Communications

Table of Contents

1. Frontmatter

2. Assistive System for the Blind with Voice Output Based on Optical Character Recognition

   D. Dhinakaran, D. Selvaraj, S. M. Udhaya Sankar, S. Pavithra, R. Boomika

   Abstract

   Everyone deserves to live freely, even those who are impaired. In recent decades, technology has focused on empowering disabled people to have as much control over their lives as possible. The braille system, which allows the blind to read, is now the only effective system available. However, this approach is time demanding, and it takes a long time to recognize the text. Our goal is to cut down on time it takes to read. Our article created a ground-breaking interactive book reader for blind people based on optical character recognition. In artificial intelligence and recognition of patterns, among the most effective technology applications are optical character recognition. It is necessary to have a simple content reader accessible, inexpensive, and easily obtainable in public. The framework is made up of a camera-based architecture that aids blind people in reading text on labels, printed notes, and objects. Text-to-speech (TTS), OCR, image processing methods, and a synthesis module are all part of our framework. Neuro-OCR deals with incorporating a complete text read-out device suited for the visually handicapped. We used Google Tesseract as an OCR and Pico as a TTS in our work. After which, the voice output is sent to the Telegram application and noticed by the user.

3. Enterprising for a Sustainable Supply Chain of Livestock and Products of Sheep Husbandry in Jammu and Kashmir

   Nadeem Younus Zargar, Nilesh Arora

   Abstract

   Sheep livestock and products are conducive for the economic growth of Jammu and Kashmir. This sector acts as not only a major source of income to the people living in the region but in terms of employment as well. This occupation requires the farmers to execute every step of the supply chain process of sheep livestock farming, including rearing of livestock to delivering the end product to the end consumer. This supply chain management process of sheep livestock and products involves traditional and unorganized techniques, which are sometimes unsustainable in ecological as well as economic context. Thus, the current study addresses the need for analyzing the impact of sustainable practices in the sheep husbandry supply chain within the region especially stressing on maintaining a balance between ecological and economic paradigms in the region. This study is required to identify areas of the traditional supply chain that can be fortified with interventional strategy of introducing modern technology and management practices. The study aims at achieving a leverage for research with the point of view of organizing sheep husbandry sector of Jammu and Kashmir as a thriving industry with potential to plug the gap between domestic demand and supply and also looks further toward becoming a hub of export of livestock and products of sheep husbandry.

4. Comparative Analysis of Object Detection Models for the Detection of Multiple Face Masks

   Saakshi Kapoor, Mukesh Kumar, Manisha Kaushal

   Abstract

   Deep learning has immense prospective in many real-life practices, one of them being object detection. Object detection based on deep learning has shown encouraging results. Since December 2019, deadly virus named CORONA or COVID-19 started to engulf the whole planet with its impact. One of the easiest and simplest ways to protect oneself from this virus is by wearing a mask. In order to detect whether a person is wearing mask or not, we propose a model to detect various face masks that include cloth masks, N-95 masks, medical masks, and no mask. The proposed model consists of two major components—annotating, labeling images and detection of face masks. A new dataset has been created by combining images from Medical Masks Dataset and Google Images, and then these images were annotated according to the mentioned categories. A comparative study has been presented among different object detection algorithms along with a proposed detection algorithm. Results show that YOLOv5 performs best in the detection of face masks when compared to other detection models. It achieved a mAP of 0.51 in just 0.24 h on our dataset. On comparing YOLOv5 to the proposed model, we found that our model achieved a precision of 0.9 as compared to 0.88 of YOLOv5. Among existing approaches YOLOv5 performed the best with precision of 0.88. The model proposed in the work results in precision of 0.90 outperforming all existing models.

5. ASL Real-Time Translator

   Pranshul Aggarwal, Kunal Kushwaha, Kush Goyal, Pooja Gupta

   Abstract

   Real-time ASL-translator would be a video conferencing application that will detect American Sign Language and will convert it into English text. In today’s scenario after the pandemic, there is a need to understand the gestures while communicating using video conferencing. As most of the existing projects work only on predicting the individual letters and not the complete words, it was quite challenging. In this work, real-time ASL-translator has been proposed and implemented that would provide a platform for video conferencing in real-time scenarios. For the same, the CNN model has been deployed on the server side. The proposed solution results with 95% of accuracy.

6. House Price Forecasting by Implementing Machine Learning Algorithms: A Comparative Study

   Ishan Joshi, Pooja Mudgil, Arpit Bisht

   Abstract

   Discerning property value via state-of-the-art machine learning techniques can evolve the current real-estate market and expose it to the technological frontiers of the modern world. This can potentially have sanguine domino effects such as opening the market to new investors as a result of technically backed price values. The current research paper strives to capitalize on this opportunity by analyzing information and data from an existing online marketplace for buyers and sellers in this industry. It is conjectured that precise prediction of house prices in a particular location through data analytics will create a candid market where prices are not arbitrary, ensuring openness in the market through P2P opportunities which will eliminate middleman charges. The research ventures to extrapolate machine learning techniques to create a model that predicts house prices in Bangalore using a plethora of algorithms such as linear regression, bagging classifier, K-nearest neighbour, XGB, decision tree, gradient boosting, and random forest. An incremental approach is deployed to gather and streamline data, clean, visualize, model and evaluate the models produced. The research is completed with a result from the comparative study, showing the most appropriate algorithm for the given data available is the random forest algorithm.

7. Comparative Study of Graph Theory for Network System

   Rajshree Dahal, Debabrata Samanta, Marimuthu Karuppiah, Jayanta Biswas

   Abstract

   The historical background of how graph theory emerged into world and gradually gained importance in different fields of study is very well stated in many books and articles. Some of the most important applications of graph theory can be seen in the field network theory. Its significance can be seen in some of the complex network systems in the field of biological system, ecological system, social systems as well as technological systems. In this paper, the basic concepts of graph theory in terms of network theory have been provided. The various network models like star network model, ring network model, and mesh network model have been presented along with their graphical representation. We have tried to establish the link between the models with the existing concepts in graph theory. Also, many application-based examples that links graph theory with network theory have been looked upon.

8. Numerical Simulation and Design of Improved Filter Bank Multiple Carrier System as Potential Waveform for 5G Communication System

   Mala Lakhwani, Kirti Vyas

   Abstract

   Single-carrier frequency division multiple access is paired with filter bank multi-carrier (FBMC) offset quadrature amplitude modulation (SC-FDMA). To improve upon the classic FBMC precoding scheme, we use a trimmed discrete fourier transform (DFT) in conjunction with one-tap scaling. SC-FDMA requires a cyclic prefix, but the proposed technique has the same peak-to-average power ratio and produces substantially less out-of-band radiation. Complex orthogonal restoration and FBMC ramp-up/ramp-down times are greatly decreased, making it possible for low-latency transmissions to be achieved. Our approach has only twice the computational complexity of pure SC-FDMA. Our assertions are backed up by simulations on channels with two levels of selection and a free MATLAB code. Note that a modified SC-FDMA transmission method can be viewed as a DFT-spread FBMC that has been trimmed. Traditional FBMC systems have more requirements for the filter, while the prototype filter has less of them. Software such as MATLAB is used throughout the entire endeavour. The proposed approach drastically reduced the PAPR of the FBMC technique by 25%.

9. Automatic Classification and Enumeration of Bacteria Cells Using Image Analysis

   Mangala Shetty, Spoorthi B. Shetty

   Abstract

   Scanning electron microscopic image processing methods are gaining considerable attention in the field of microbiology. It is critical to classify and enumerate accurately the population of microbes in the preparation and express this information to the consumer on the product label. The other crucial application of classification and counting of microbial cell is in the field of medical microbiology to search and detect the causes of diseases. The manual process of bacteria cell inspection is tiresome and eye-straining and depends on the experience of the individual in the laboratory experimenting the bacterial samples, and it is a time-consuming procedure. Various experiments have been made for the replacement of manual observation by automatic inspection of microbiological information. One of the significant projects in this direction is the scanning electron microscopic (SEM) bacteria cell image analysis. Extracting knowledge from image information is a difficult job in biological image processing. This paper proposes a fully automated classification and counting system for lactic acid bacterial cell using image processing methods based on marker-controlled watershed method. Proposed technique ultimately will strengthen the accuracy and reliability of probiotic strain enumeration and classification.

10. Liver Cirrhosis Stage Prediction Using Machine Learning: Multiclass Classification

    Tejasv Singh Sidana, Saransh Singhal, Shruti Gupta, Ruchi Goel

    Abstract

    Liver cirrhosis is a disease that affects a large population worldwide. Liver cirrhosis is further divided into four stages. This paper aims to predict the stage of liver cirrhosis of a patient using machine learning. It is a supervised learning problem of multiclass classification. Seven different algorithms were used for this purpose, and their performance was analyzed and compared in order to find the best approach. Different scaling and feature selection strategies were used in order to study their effect on the performance of various algorithms. It was found that an ANN-based approach achieves the best performance for this particular problem. A feature selection approach based on random forest and mutual information (RF + MI) was proposed in this paper, and its performance was compared with the standard Random Forest (RF) method for feature selection in classification problems. Experimental results demonstrated that the RF + MI approach shows minor improvement in comparison with random Forest (RF) for feature selection.

11. Dynamic State Estimation of a Multi-source Isolated Power System Using Unscented Kalman Filter

    Neha Aggarwal, Aparna N. Mahajan, Neelu Nagpal

    Abstract

    In power systems, dynamic state estimation (DSE) is a crucial activity for real-time monitoring and control to ensure the system’s safe and efficient operation. This paper presents an method for real-time estimation of dynamic states of an isolated power system integrated with renewable energy sources (RESs) and electric vehicles (EVs) aggregates. The proposed method employs an adapted unscented Kalman filter (UKF) as an observer to estimate the system’s dynamic states which are either inaccessible or corrupted with measurement noise. MATLAB/Simulink is used to develop a simulation platform for frequency response model of power system. The simulation results on the developed test system investigated the efficacy of UKF as dynamic state estimator that takes into account the diverse behaviours of the system and provides accurate estimates of the system states.

12. Investigating Part-of-Speech Tagging in Khasi Using Naïve Bayes and Support Vector Machine

    Sunita Warjri, Partha Pakray, Saralin A. Lyngdoh, Arnab Kumar Maji

    Abstract

    This paper presents the investigation of the Khasi language toward the PoS tagging systems. Khasi is an Austroasiatic language, which is spoken in Meghalaya, India. The foremost purpose of this paper is to develop part-of-speech (PoS) tagging for the Khasi language based on the support vector machine (SVM) and the Naïve Bayes (NB). This work is the first instance using SVM and NB approaches in Khasi for PoS tagging. Part-of-speech tagging performs a vital role in natural language processing (NLP). In this research work, we have used a PoS tagging corpus that is manually tagged by using the designed grammatical PoS classes. The annotation is done using 53 tags, and the corpus consists of around 75,000 tokens. The PoS tagging system is trained and tested with a ratio of 80:20. It is found that the system yielded promising outcomes while comparing the state-of-art results.

13. Machine Learning and Deep Learning-Based Detection and Analysis of COVID-19 in Chest X-Ray Images

    Kunal Kumar, Harsh Shokeen, Shalini Gambhir, Ashwani Kumar, Amar Saraswat

    Abstract

    Machine learning (ML) is a cutting-edge method with numerous applications in prediction and classification. This technology should be used to identify high-risk patients, their death rates and other irregularities in the COVID-19 pandemic (Taresh et al. in Int J Biomed Imaging, 2021 [1]). ML can be used to learn more about the virus’s nature and to foresee potential problems. With the goal in mind to help the healthcare sector, we can definitely leverage the advancement of technology (Chowdhury et al. in IEEE Access 8:132665–132676, 2020 [2]). This paper uses the COVID-19 dataset available on Kaggle. Various machine learning techniques are used to weigh the risk of COVID-19 disease in a patient in the proposed work. VGG19, MobileNetV2, DenseNet201, CapsNet201, COVID-Net, CoroNet and VGG16 are tested for classifying the images of normal human lungs versus lungs affected by viral pneumonia due to COVID-19. The performance of various machine learning algorithms is analysed, and it was determined that VGG16 algorithm achieved the best accuracy (97%) in tests.

14. A Comprehensive Study of Machine Learning Techniques for Diabetic Retinopathy Detection

    Rachna Kumari, Sanjeev Kumar, Sunila Godara

    Abstract

    Diabetic retinopathy is a threatening complication of diabetes, occurred due to damaged blood vessels of light-sensitive areas of the retina. DR leads to total or partial blindness if left untreated. DR does not give any symptoms at early stages, so earlier detection of DR is a big challenge for proper treatment of diseases. With advancement of technology various machine learning approaches based methods are designed for early detection of DR so that experts can provide proper treatment to the patients for preventing its harmful effects. This paper provides a comprehensive study of machine learning based approaches, e.g., Bossa Nova feature beyond lesson, pixel-based super classification, SVM and Gaussian Mixture Model, selective sampling and Patch-based sampling, Carl Zesis Meditec ML, BoVW Salient map, U-Net, LeNet, and STSF deep architecture, etc., used to detect diabetic retinopathy.

15. Evolution of WSN into WSN-IoT: A Study on its Architecture and Integration Challenges

    Radhika Dhiman, Jawahar Thakur

    Abstract

    Researchers have contributed a lot to the enhancement of wireless sensor networks (WSN). Various protocols have been designed for these networks to work with limited resources. However, we are leading toward a new era of innovation, and a new network of things (devices), known as Internet-of-Things (IoT), is evolving, where everything will be connected to the Internet. WSNs integration with the Internet will unleash the full potential of sensor networks, and the sensed data will be available to any user on the Internet at any time. However, their integration raises some challenges which need to be tackled. WSN protocols, designed for limited resources, may not be compatible to create a robust connection with the Internet. Also, the connectivity will make WSN accessible to the whole Internet that may influence the working of the sensor networks. In this paper, we have reviewed the evolution of WSN toward WSN-IoT. For this, we have studied the layered architecture of WSN and how various researchers have contributed to its lifetime enhancement. After that, we have compared the architecture of WSN with IoT to find out the architectural dissimilarities between them. Finally, to address the various challenges that emerge with WSN-IoT integration, we have reviewed the different approaches used for their integration, and then some solutions are given to deal with those challenges.

16. Big Data Security Trends

    Reenu Bhatia, Manu Sood

    Abstract

    The continuous tremendous growth in big data gives birth to many associated issues such as efficient/effective handling, processing, securing, maintaining privacy, and managing transaction logs. As growth cannot remain confined only to the efficient handling of such data, the hackers too are following hot pursuits to develop equally efficient or even better mowers which are sufficient enough to penetrate all the security walls of such data. With the traditional security solutions it is very difficult to build a security framework which can handle these types of attacks. In this paper, we discussed the traditional as well as modern security solutions to the big data security problem. Here we discuss the current issues, their solutions and limitations of the existing solution. The aim of this work is to provide a detailed review of the latest big data security trends.

17. Application of NLP and Machine Learning for Mental Health Improvement

    Trinayan Borah, S. Ganesh Kumar

    Abstract

    Humans’ most powerful tool is their mental wellness. Individuals’ well-being can be impacted by poor mental health. This paper focuses on a smart technical solution to the problem of mental health issues detection related to the stress, sadness, depression, anxiety, etc. which if not handled efficiently may further lead to a severe problem. The paper deals with the designing of an automated smart system using social media posts that will help mental health experts to successfully identify and understand about the mental health condition of social media users. That can be done based on text analysis of rich social media resources such as Reddit, Twitter posts. The implementation of the system is done using Natural Language Processing (NLP) methods, machine learning and deep learning algorithms. The models are trained using a prepared dataset of social media postings. With this automated system the mental health experts can able to detect the stress or some other emotions of social media uses in a very earlier as well as faster way. The proposed system can predict five emotional categories: ‘Happy’, ‘Angry’, ‘Surprise’, ‘Sad’, ‘Fear’ based on machine learning (Logistic Regression, Random Forest, SVM), deep learning Long Short-Term Memory (LSTM) and BERT transfer learning algorithms. All the applied algorithms are evaluated using confusion matrix, the highest accuracy and f1 score achieved is more than 85%, which is better than the existing human emotion detection systems.

18. Energy Efficient RPL Objective Function Using FIT IoT-Lab

    Spoorthi B. Shetty, Mangala Shetty

    Abstract

    The Internet of Things connects and establishes communication between the heterogeneous devices. The network used in IoT is low power and lossy networks which is commonly known as LLN. The components in LLN use low power for its operations. The Internet Engineering Task Force (IETF) has defined routing protocol for standardized LLN, i.e., routing protocol for low power and lossy networks (RPL). The RPL selects suitable path to reach the destination by constructing the destination oriented directed acyclic graph (DODAG). The DODAG can be constructed based on the type of objective function. RPL commonly uses two objective functions they are, MRHOF and objective function zero. The objective function zero uses hop count as its metric, and MRHOF uses the metric of expected transmission count. Hence, selection of the energy efficient objective function plays an important role to make the network of IoT more energy efficient. In existing research, the superiority of MRHOF and OF0 is experimented using only simulation not using the real testbed. Hence, it is essential to conduct the experiment in real testbed to identify the energy efficient objective function. Future Internet of Things (FIT) lab is a platform to carry out the large-scale experimentation using testbed. In this paper, two objective functions MRHOF and OF0 are considered, and experiments are performed in FIT IoT-Lab to identify the energy efficient objective function. After the experiment, it is identified that objective function zero is more energy efficient than MRHOF. From the results of simulation and real testbed experiments, it is concluded that OF0 consumes less energy than the MRHOF.