Micro-Electronics and Telecommunication Engineering

Table of Contents

1. An Intelligent Wheelchair for Disabled Person

   S. Prabhu, V. Nagaraju, P. Sailaja, B. R. Tapas Bapu

   Abstract

   Every new development in sensors and ultrasonic technologies has always been a bonus for electronics traveling aids (ETAs). This sets way to the development of smart vehicles for disabled people. These smart vehicles are used by physically disabled people. This electronic intelligent wheelchair is designed to help disabled people to move independently to any place without a caretaker. It consists of a smartphone app-based operation along with obstacle avoidance, fall detection and narrow way locomotion. An android application is used to control the wheelchair. This smart chair can function in full automatic mode. It uses GPS-guided system and magnetic compass to determine location and stores the location to return back to the location automatically. The fall detection system tilts the seat when it moves around any inclined plane and alerts the nearby caretaker with an SMS when a fall is detected. It has an intelligent collision avoidance system to avoid any type of obstacles including potholes and can move around narrow pathways and doors without being hit or bumped against any obstacle.

2. Convolutional Neural Network-Based Detection and Classification of Cardiovascular Disease and Diabetic Macular Edema

   R. Senthamizh Selvi, S. Bragadesh Bharatwaj, B. Ajith Kumar, V. R. Bharath Raj, S. Sudha

   Abstract

   Artificial intelligence is one of the most advanced and prominent research areas which is used to solve several worldly problems. Traditional methods take up a lot of the doctor’s time and are also prone to human error. The existing techniques utilize separate algorithms to only detect the disease. But their speed and accuracy are very low. The developed novel system utilizes deep learning in which the retinal fundus images are first preprocessed and then sent to a convolutional neural network (CNN) module. The proposed module not only accurately detects the presence of a disease (i.e., diabetic macular edema (DME) or cardiovascular disease) but also determines its severity. The severity is determined by the intensities of the pixels in the eye fundus images. Based on the empirical results obtained on the parameters, accuracy, sensitivity, and specificity, the proposed CNN model is proven to be more efficient and accurate than the previously utilized algorithms such as support vector machine (SVM) and extreme learning machine (ELM).

3. Classification and Detection of Malarial Parasite in Blood Samples Using K-Means Clustering Algorithm and Support Vector Machine Classifier

   R. Mohana Priya, L. K. Hema, V. Vanitha, R. Karthikeyan

   Abstract

   The natural history of malaria involves cyclical infection of humans and female Anopheles mosquitoes. In humans, the parasites grow and multiply first in the liver cells and then in the red cells of the blood. The early diagnosis of malaria is required; otherwise, it leads to death. In this study, an effective method for the classification and segmentation of malaria parasite using k-means clustering (KMC) segmentation algorithm and support vector machine (SVM) classifier is presented. Initially, the input blood sample images are given to KMC segmentation technique for segmentation. Then the segmented image is given to statistical features like mean and standard deviation for feature extraction. Then the extracted features are saved in the feature database and used for classification using SVM classifier. The classification of healthy and affected cells of malaria parasite in blood sample images is made by using SVM classifier. Experimental result shows the performance of the proposed system.

4. FPGA Implementation of Parallel Adder Using Reversible Logic Gates

   S. A. Yuvaraj, K. Gunasekaran, D. Muthukumaran, K. Umapathy

   Abstract

   Reversible digital technology can now start taking a more desirable direction for low dissipation of power, higher processing speeds. Here, we suggested the construction of an 8-, 16-, 32-, 64-bit multiplier using the carry-save adder, the Kogge stone adder, and the HNFG adder with the high operating speed of the proposed HNFG gate adder. The architecture of the device and the logic gates which are reversible can be implemented using the Vedic multiplier. The output of the accumulator operation is dependent on the multiplier unit and the adder units. Here, the development of a multiplier and an adder can be built using reversible gates to achieve high operating speeds, and the use of a Vedic multiplier for greater efficiency. Also fewer area and partial elements are used. A comparative study is being conducted between the conventional [1] and the Kogge stone adder based on HNFG. Finally, it has been shown that the proposed HNFG gate with multiplier adder has a relatively high-speed over the conventional method. The whole process of simulation and synthesis is done using Xilinx 14.7 tool.

5. Smart Sentimental Analysis of the Impact of Social Media on COVID-19

   A. Mugilan, R. Kanmani, M. Deva Priya, A. Christy Jeba Malar, R. Suganya

   Abstract

   Coronavirus or COVID-19 pandemic has spread to 210 countries and territories taking the lives of more than 140,000 people globally as per the record on April 2020. This worldwide outburst is a predominant topic of discussion on the social media. This paper investigates the Twitter data to analyze the sentiment of the public regarding this pandemic. Long Short-Term Memory (LSTM) algorithm-based Recurrent Neural Network (RNN) architecture is implemented for performing sentiment analysis. Further, data related to coronavirus are collected from social media sources such as Blogs, Forum, News, Videos, Web, Podcast using hashtags related to coronavirus. Data from 15 March 2020 to 23 March 2020 are used for analysis. From this analysis, it is found that the information related to corona virus has considerably influenced the social media. People are well aware of this outbreak. Analysis has also shown that there is less spread of false information regarding coronavirus unlike SARS, MERS, etc.

6. Dermoscopic Image Classification Using Two-Stage Processing of Shearlet Features with Support Vector Machine

   S. Mohan Kumar, T. Kumanan

   Abstract

   Automated classification of dermoscopic images is of importance for the interpretation of skin cancer. In order to avoid false detection, image classification is used in medical field for detecting abnormality. In this paper, a system is developed based on shearlet and support vector machine (SVM) that will assist the automated interpretation of dermoscopic images for cancer diagnosis. This system provides an efficient way of classifying dermoscopic images that can be used as a second reader. It consists of two main stages: local processing stage (LPS) and global processing stage (GPS). In LPS, the energy features from different levels and directions are individually analyzed using SVM classifier. In GPS, the best directional features from each level are combined, and the classification accuracy is improved by selecting more dominant features by a statistical approach. Results show that the features from different level improve the classification yield for SVM classifier from 96% (LPS) to 99.5% (GPS) on PH² database images.

7. Delay and Energy-Aware Forwarder Selection in Wireless Sensor Network

   Vamsidhar Yendapalli, B. Ramesh Naik

   Abstract

   Wireless sensor network (WSN) is a promising technology which alters data retrieval from the surroundings through heavily distributed small, least cost as well as less energy wireless device known as sensor nodes. This paper proposes a delay- and energy-aware forwarder (DEAF) selection in WSN. This strategy chooses a route with the least cost that is expected delay (ED) in a multi-hop WSN. Here, the expected delay is computed by queuing delay as well as communication delay via wireless connections. The forwarder node selected by less expected delay as well as the node has the highest amount of outstanding energy. Thus, the source transmits the data packet without packet losses in the network. Simulation is carried out in the network using network simulator-2 and is known that, the introduced scheme performs superior packet obtained rate and lesser delay than the conventional scheme.

8. An Effective Approach for Integrating Microsoft Power BI Application with Python for Predictive Analytics

   Vipul Vashisht, Nidhi Jakhmola, Pritish Manjarwar, Nirav Nikhil

   Abstract

   With the advent of technology and interest of organizations in artificial intelligence (AI), there has been enormous focus in implementing prediction algorithms in data-centric organizations for futuristic business planning. So far, very few approaches have been proposed for integrating machine learning with business intelligence tools, and due to lack of technical documentation, most of these approaches have limited adoption in practice. The objective of this paper is to offer an easy to understand approach that could be quickly adopted for integrating the business intelligence (BI) tool with machine learning (ML). This communication describes a process of applying artificial intelligence technologies, in particular python programming language for integration with Power BI, one of the mostly used business intelligence tools from Microsoft.

9. Traffic Sign Recognition Using Multi-layer Color Texture and Shape Feature Based on Neural Network Classifier

   Manisha Vashisht, Brijesh Kumar

   Abstract

   In the last decades, with the rapid changes taking place in both hardware and software technology domains, the field of image acquisition, image detection, and recognition has undergone drastic revolution bringing innovations for overall improvements. So far, there has been limited research conducted on mapillary dataset of traffic signage images, and the results obtained are not so encouraging. The research contributes toward providing an innovative approach of extracting color features and then using artificial neural network for image recognition purposes. This research is conducted on a mapillary image dataset of 5000 images, and the results have found to be encouraging. The dataset has been made freely available for academic research.

10. Secured Library Access Through Face Recognition Integrated with RFID Technology

    P. Anantha Prabha, S. A. Vighneshbalaji, M. Deva Priya, R. Suguna Devi

    Abstract

    Securing resources in the library is one of the prime concerns by preventing the entry of unauthorized users into the library. Saving time of users at entry/exit and circulation points is the next key concern. To access the library, the member is expected to swipe or scan his smart or IDentity (ID) card at the entry/exit point, and he/she is expected to wait at circulation points for the issue of books. In this system, members cannot access the library if they forgot to bring or miss their cards. Unauthorized persons can also easily enter into the library with stolen or duplicate ID cards. In order to secure and automate the library, a new system of accessing the library is proposed with deep learning and Radio-Frequency IDentification (RFID) technology. The proposed system focuses on automating the authentication of user with face recognition at the entry/exit point (turnstile) and providing them good user experience and automating the issue of books (RFID tagged) by using RFID detectors at the exit point. This system yields good results with our test datasets. Activation of this system at library premises ensures security and also facilitates auto checks-in and out, and thus saves users' valuable time.

11. Real-Time Human Body Tracking System for Posture and Movement Using Skeleton-Based Segmentation

    Venkatesh Gauri Shankar, Bali Devi, Ujwal Sachdeva, Harsh Harsola

    Abstract

    Real-time tracking of the human body has proven to be a key technology in several areas. In our proposed model, we have initially performed segmentation of bodies from a live feed and further detected and classified all the body parts to create a skeleton which does analysis of the posture and movement of the body. As this model works on a live feed, it has multiple real-time applications which include, but are not restricted to, applications in the field of medical diagnostics, provision of improved safety and security, human–computer interface, etc. There have been multiple instances and various research works as cited later in this paper that has attempted to track the human body. Myriad different solutions have been presented using different sensors attached at joints for estimation of the poses, but these pose estimations systems are quite personalized and have huge setup costs. The system makes a good tradeoff between system simplicity and accuracy and helps to distinctly identify and track human bodies in real time.

12. Enhanced InGaAs/InAs/InGaAs Composite Channel MOSHEMT Device Performance by Using Double Gate Recessed Structure with HfO2 as Dielectric Materials

    R. Saravana Kumar, N. Mohankumar, S. Baskaran, R. Poornachandran

    Abstract

    This work reports that the composite channel InGaAs, InAs and InGaAs thin, with dual delta-doped double recessed gate (DG) MOSHEMT, is 40 nm gate length, the barrier 3 nm and 15 nm thick channel, and this structure has been simulated in the TCAD Sentaurus simulation tool. The DC and RF characteristics of proposed device are characterized by different gate lengths along with different V_DS. The novel design features included under this proposed structure, namely recessed high stem gate, thin barrier, dual gate, composite channel and the HfO₂ as a dielectric material, are applicable for low leakage current along with Tera Hertz frequency applications. The simulation results show the exhibition of a high drain current of 1.38 × 10⁻³ A/µm, transconductance of 3.35 × 10⁻³ S/µm, threshold voltage of 0.13 V, cutoff and maximum frequency of oscillation of 791 and 995 GHz by the DGMOSHEMT devices at L_G = 40 nm and V_DS = 0.5 V. The findings are obtained  due to the superior electron transportation properties of a DG MOSHEMT structure with compound semiconductor III-V materials.

13. FPGA Realization of Reconfigurable DA-Based Digital FIR Filter Using DRPPG and MCSA Techniques

    G. O. Jijina, R. Mohana Priya, P. Solainayagi

    Abstract

    This article presents an efficient multiplication and accumulation (MAC) approaches called “distributed arithmetic (DA) based MAC” and “reconfigurable implementation-based MAC” for alleviating the hardware performances of digital filters. Both reconfigurable implementation and DA-based techniques are combined in this paper at the first time with the help of suitable accumulation structure called “modified carry-save adder (MCSA)” and parallel reconfigurable structure called “dynamically reconfigurable architecture (DRA). Traditionally, DA-based FIR filter implementations require large size of look-up table (LUT) for storing the filter coefficient values. The most disadvantage of DA-based FIR filter implementation is the absence of configurability. Hence, traditional DA-based FIR filter is not suited to wider range of applications. In order to reduce the problem, reconfigurable technique is introduced in MAC unit of DA-based FIR filter implementation. To reduce the complexity form of partial product generation, Reduced Wallace Tree Generation (RWTG) is used along with MAC unit of reconfigurable DA-based FIR filter implementation. In the final stage of RWTG, efficient adder structures are essential to add “n” bit binary data. To meet this requirement, MCSA-based adder is used in the final stage of RWTG. Finally, dynamic or distributed reconfigurable partial product generation (DRRPG) is introduced for producing parallel execution of PPG. Proposed design is estimated by using ALTERA field programmable gate array (FPGA) board by using Quartus II Web Edition tool. The performances of proposed design are validated after implementing in ALTERA FPGA design tool.

14. Indian Air Quality Health Index Analysis Using Exploratory Data Analysis

    Venkatesh Gauri Shankar, Bali Devi, Anurag Bhatnagar, Akhilesh Kumar Sharma, Devesh Kumar Srivastava

    Abstract

    Air quality management requires dependable data of air quality gathered, assessed and broke down normally. This is of foremost significance in shielding man and his condition from harming introduction to air contamination. The information for SPM, SO₂ and NO₂ were gathered at three locales speaking to private, business and mechanical action zones at all the destinations. In light of this information, the Air Quality Record (AQI) was determined utilizing Oak Edge Air Quality File (ORAQI). This means that the absolute impact of the considerable number of contaminations together. Figuring’s of AQI for various seasons and diverse movement zones are done to look at the contamination level. In the current article an endeavour is likewise made to consider the adjustment in contamination status during the most recent decade utilizing the month to month mean information. To investigate and anticipate the degree of Air File level throughout the years from 1990 to 2015 to show signs of improvement comprehension of the patterns and examine how significant government strategies have influenced the Quality File. The point of this paper is to advise open with respect to by and large prominence of air quality through a rundown boundary that is straightforward, Educate residents about related wellbeing effects of air contamination presentation; and Rank urban areas/towns for organizing activities dependent on proportion of AQI.

15. Racist Tweets-based Sentiment Analysis Using Individual and Ensemble Classifiers

    Bali Devi, Venkatesh Gauri Shankar, Sumit Srivastava, Kriti Nigam, Lakshay Narang

    Abstract

    In this time of a global pandemic, as our society is pushing forward to a more digital culture of working, living, studying, and earning, the Internet’s significance is unprecedented, and inevitably so, the use of various social media platforms has seen a boom. This paper focused on hate speech racism and sexism and built a refined classifier that detects racist and sexist comments from the tweets. Investigating accessible data from Internet-based existence may produce interesting findings and bits of information about virtually every object, organization, or individual inside the universe of general feelings. Opinion mining is the statistical action behavior of suppositions, notions, and subjectivity of content from a corpus that integrates natural language processing (NLP) and machine learning (ML) to classify a range of emotions. As such, sentiment analysis remains a widely researched and ever-evolving topic in the field of NLP. XGBoost with word2vec gave us the most promising results, so we refined it further by fine-tuning. Furthermore, we observed that the F1 score was 0.690285. Thus, we were able to reach an accuracy of 69% in our classifier.

16. An Efficient Model for High Availability Data in Hadoop 1.2.1

    Anurag Bhatnagar, Venkatesh Gauri Shankar, Bali Devi, Nikhar Bhatnagar

    Abstract

    Day by day the data is increasing enormously so storage of big data and analysis of big data require large storage as well as fast processing. Scaling can be done in two ways horizontal scaling and vertical scaling. Earlier horizontal scaling was approached to process big data. They tend to increase processing power each time the data is increased. Today, the vertical scaling approach is the best practice to handle big data. Privacy is also major concern, public fear of inapt use of their personal data.Apache Hadoop is one of the solution of big data. Hadoop below version 2 lack high availability in NameNode. We have found the solution of high availability of NameNode in Hadoop version 1. We have used centralized storage that is NFS server and shared that storage with two places. We have called it primary and secondary NameNode according to the vocabulary of Hadoop version 2. Primary NameNode is working as a master to the DataNodes while secondary NameNode will keep track of live status of primary NameNode. When primary NameNode goes down then secondary NameNode will replace its IP with primary NameNode’s IP and then it will become master to the DataNodes. In this way, we can obtain high availability in older Hadoop version older than 2.

17. A Framework for Detection and Validation of Fake News via Authorize Source Matching

    Deepak Mangal, Dilip Kumar Sharma

    Abstract

    In the present era, peoples are sharing views, information, and knowledge on social media across the world without validating the contents. This increases the probability that deceptive news reaches the group of peoples. This type of deceptive news is termed as fake news and requires a proper solution to validate such contents. To overcome and find a solution, a novel framework along with algorithm has been proposed in this research work. The framework is using embedded image/text as input. Extract text with image features used as query and sent to multiple search engines to find relevant links to validate source of generation. After all, the source of the selected links validates by stores authorize source list. Algorithm achieves 82, 85, and 94% accuracy on MediaEval, BuzzFeedNews datasets, and Google news, respectively.

18. Prediction of Type of Bone Disease Using Machine Learning

    Sujata Joshi, Mydhili K. Nair

    Abstract

    Machine learning has emerged as a promising field in the area of health care because of its ability to automatically learn and improve from experience without explicit programming and thus can discover useful patterns. The discovered patterns are then used for further analysis and decision making and making predictions. Bone is an integral part of human body which provides shape and structure to the body. As bones are used throughout the lifespan of a person, it is quite possible that bones get diseased or injured. The task of identifying the type of bone problem is quite a challenging task as it requires years of experience for healthcare professionals and intense medical tests, X-rays and scanning to be conducted. The number of such practitioners available is limited. Also, the bone ailment is quite painful and the patients have long waiting times. Hence, there is a requirement to detect the type of bone disease or ailment, quickly and accurately. The objective of this work is to develop a predictive model to identify type of bone disease from physical examination features. The unique features of this work is that we have used live data of patients to develop predictive models. We have proposed an algorithm which uses a combination of k-NN learning algorithm and Jaccard similarity measure for the purpose of prediction. The accuracy and error rate of the model developed by the proposed method is found to be 80.5% and 19.5%, respectively.