main-content

Über dieses Buch

This book presents the select peer-reviewed proceedings of the International Conference on Signal and Data Processing (ICSDP) 2019. It examines and deliberates on the recent progresses in the areas of communication and signal processing. The book includes topics on the recent advances in the areas of wired and wireless communication, low complexity architecture of MIMO receivers, applications on wireless sensor networks and internet of things, signal processing, image processing and computer vision, VLSI embedded systems, cognitive networks, power electronics and automation, mechatronics based applications, systems and control, cognitive science and machine intelligence, information security and big data. The contents of this book will be useful for beginners, researchers, and professionals interested in the area of communication, signal processing, and allied fields.

Inhaltsverzeichnis

Deep Semantic Segmentation for Self-driving Cars

A self-driving car comprises three subsystems in the navigation, viz. lane finding, urban scene understanding and geopositioning. This paper introduces the technique of semantic segmentation for urban scene understanding with various implementations in recent years and proposes a novel approach to fast and accurate semantic segmentation. The architecture of the model is designed using VGG16 as encoder, adopting hierarchical feature fusion technique to perform downsampling followed by a lightweight decoder to perform upsampling. Skip connections are introduced between the encoder and the decoder to improve the information flow. This new approach outperforms the previous implementations by giving 94% accuracy and 8.7 frames per second (FPS) simultaneously.

Abhilash Sachin Kulkarni, Jyothi S. Nayak, Aditi Desai, Jahnavi Singh, Shraddha Murali

Trends in Energy Management System for Smart Microgrid—An Overview

Smart Grid (SG) is an integrated grid to improve efficiency, consistency, and security with the incorporation of conventional and renewable energy sources, through automated control and modern communication technology. The development of Microgrid (MG) is a great solution for the incorporation of sustainable energy resources inside the smart grid environment. The emergence of Microgrid (MG) by decomposition of the grid is a combination of Distributed Energy Resources (DERs), Energy Storage System (ESS), loads and Control devices. This makes MG a single and controllable power supply system that can enhance versatility, dependability and furthermore incorporate the benefits of distributed generation. In MG, an EMS is fundamental for the ideal utilization of distributed energy sources in smart, secure, reliable and synchronized ways. The need for understanding the energy utilized is increasing since effective Energy Management is more challenging in Microgrid (MG). This paper reviews several Energy management systems developed based on different strategic approaches available for Microgrid on Demand-Side Management.

Harini Vaikund, S. G. Srivani

IRIS: An Application for the Visually Impaired Using Google Cloud API

The visually impaired struggle to walk safely without having any human guidance. They face several problems in their lives which include identifying the hindrances when they are walking. Traveling from one place to another is a tedious task for them requiring the help of other people. White canes and guide dogs are usable but are not fully reliable. Due to this disability, they have to go through a lot of hardships to conquer over the problems in navigation. Thus there is a need for a smart solution which would overcome the challenges faced. Focus is on designing one kind of a visual aid which is cost-effective and efficient. In this context, we propose a smart stick (IRIS) to help the user in obstacle detection and navigation. A conventional PVC pipe forms the stick on which ultrasonic sensors, camera, GPS module, GSM Module which are all interfaced with Raspberry Pi are mounted at different positions to detect and identify obstacles in the path of the user. Google Cloud Vision API recognizes the objects in the surroundings and delivers an audio signal about the same via earphones. The sense of sight of the user is now achieved by the sense of hearing. It also uses a GSM Module and a GPS module to contact the user’s relatives in emergency situation by sending them the user’s location. The whole device’s design ensures independence to the user.

Kajal Naik, Nikita Sawant, Gauravi Kamat, Siddhi Kandolkar, Niyan Marchon

Shot Boundary Detection Using Artificial Neural Network

As of late video is the most utilized information type on the Internet. Content, sound, and pictures are consolidated to establish a video, so recordings are enormous in size. The human mind can process visual media quicker than it can process message. This expansion in information has required the investigation of powerful strategies to process and store information content. In this paper, we have suggested a hybrid video shot boundary detection process using feature extraction by mean log difference which is combined with artificial neural network. We devised two-step method for automatic shot boundary detection. Firstly, features are extracted using H, V, S procedure along with histogram distribution technique, and then, this mean log difference array is applied as an input to ANN which identifies video shots based on probability function. We have incorporated feed-forward network structure which processes nonlinear factual information to calculate shot boundary detection considering probability function. Finally, we have evaluated the results using precision, recall, and F1 measure. An experimental result indicates that ANN along with mean log difference, it offers efficient representation of shot boundaries and the results are satisfactory. Comparing the proposed method with improved block color feature method, there is a sort of trade-off relation between the two algorithms, and it is observed that for fast characteristic variations, ANN performs moderately better while for complex videos improved block color feature method is suited in better way.

Neelam Labhade-Kumar, Yogeshkumar Sharma, Parul S. Arora

Custard Apple Leaf Parameter Analysis, Leaf Diseases, and Nutritional Deficiencies Detection Using Machine Learning

Custard apple (Annona Squamosa L.) is the oldest fruit plant in the dry land. It is begun from a tropical area of America and widely disseminated all through the tropics and subtropics. The custard apple fruits are cultivated in many states in India on a commercial scale. Disease detection and health monitoring in a plant are essential for sustainable agriculture. Nutrients play a crucial role in influencing tree growth, fruit production, and fruit quality. It is arduous for human vision to identify the particular leaf disease and nutrient deficiency by naked eyes. In this paper, an attempt is made to propose a system for leaf parameter analysis, detection of N, P, K deficiencies, and leaf diseases. K-nearest neighbors (k-NN), and Support vector machine (SVM) algorithms are applied for the classification of leaf deficiencies and leaf diseases. Database of 125 and 80 Custard apple leaf images are used for leaf diseases and deficiencies, respectively. Experimental results showed that the proposed leaf parameter measurement system had attained 99.5% accuracy. This paper exercise a supervised machine learning approach using image processing.

Discontinuous PWM Techniques to Eliminate Over-Charging Effects in Four-Level Five-Phase Induction Machine Drives

Four-level dual inverter configuration fed five-phase induction motor drive utilizing unequal DC voltages for the dual two-level inverters on both sides of the open-ended stator is discussed in this paper. The DC voltages of the two inverters are maintained in the ratio of 2:1. By properly adding zero-sequence signals to the reference signals, continuous and discontinuous pulse width modulation (DPWM3) can be generated. But in continuous PWM technique, extensive speed control is not possible since over charging of capacitors and the DC voltage ratio of 2:1 fail to maintain. To solve this problem, an off-set value addition to the reference modulating signal is implemented. Compared to continuous PWM technique, in DPWM3 techniques, a better improvement in the quality of output voltage and current is observed and is discussed in terms of harmonic order; number of levels in common mode voltage (CMV), total harmonic distortion (THD). The proposed techniques are designed using a simple, easier, scalar-based approach which depends upon instantaneous values of the reference voltages and it does not need complex mathematical calculations. Simulation and experimental studies have been carried to show the efficient performance of the drive.

J. Balakrishna, Teegala Bramhananda Reddy, Marapu Vijaya Kumar

State of Charge Estimation Using Extended Kalman Filter

As the Automobile Industry is currently focusing on the development of Electric vehicles, battery engineering has taken a boon. State of Charge is the ratio of the amount of charge available in the battery to the rated charge capacity of the battery. This paper presents equivalent circuit modeling of a Li-ion battery cell and its state of charge estimation using the Kalman Filter algorithm in MATLAB Simulink. The simulation result indicates an error of about 1% in estimation.

Frequency and Pattern Reconfigurable Antenna for WLAN and WiMAX Application

An innovative bow tie frequency and pattern reconfigurable antenna is proposed in this paper which can be used for WLAN and WiMAX applications. The proposed antenna is a modified bow tie antenna with an additional branch on patch side. Two p-i-n junction (PIN) diodes are used for obtaining the required frequency and radiation pattern. One diode is placed on the patch side and another diode is placed on the ground side. The antenna operates at three different frequency values 3.35, 3.6 and 5.2 GHz with changing radiation patterns for each operating frequency. During simulation maximum efficiency of 80% along with a maximum gain of 1.2 dBi is obtained for on switch condition and an efficiency of 91% and gain of 2.1 dBi is obtained for off switch condition.

Anuradha A. Palsokar, Swapnil L. Lahudkar

Implementation and Analysis of Low Power Consumption Full Swing GDI Full Adders

Power efficiency of any design can be obtained in terms of PDP. The approach used to design any system defines the performance of system. Here, Gate Diffusion Input (GDI) design techniques as well as CMOS design techniques are used for designing full adder circuits. Full swing GDI technique is utilized to reduce power consumption and delay. Full swing GDI technique gives better speed of operation as compared to CMOS technique. While keeping these parameters at best GDI maintains low complexity of design. Full adder circuits are designed using 180 nm technology node in Cadence Virtuoso with supply voltage of 1.8 V.

Single Image Rain Removal Using Convolutional Neural Network

The visibility of the image is highly affected by the rain streaks. This can influence the performance of numerous visual tasks, for example, image enhancement, object tracking, recognition, surveillance and autonomous navigation. Process of recognition and removal of rain streaks is a quite complex and difficult task since there is no spatial-temporal content of rain streaks in a single image as compared to the dynamic video. This paper proposes an improved convolutional neural network (CNN) architecture to recognize and remove the rain streaks. Linear additive composite model is used for making rainy image model. Network is trained on the pre-processed image, which helps to enhance the learning of the network weights and training without huge increase in training data or computational resources. The experimental work shows that the CNN architecture successfully performed on both synthesized and real-world rainy images.

P. Musafira, K. S. Shanthini

Ring Oscillator-Based Physical Unclonable Functions

Physical unclonable function is playing an important and efficient role in system security. Ring oscillator is basically a delay-based PUF, and during fabrication process variations, the delay introduced is used for detecting secrecy of the PUF design. A framework of ring-oscillator PUF is built to check the unpredictability of the response based on the challenge created by a 4 bit-LFSRs. Simulation results of ring-oscillator PUF show that the response bits generated are unique for every challenge. Experimental results of National Institute of Standard technology Test (NIST) Suite demonstrate that the PUF’s secrecy generated by the ring-oscillator PUF is random, and it varies from different FPGA platforms. The ring-oscillator PUF is evaluated according to the metrics namely security, uniqueness, and randomness of the response bits generated. The RO-PUF uniqueness and randomness calculated are more efficient in comparison with any other RO-PUF implemented.

Shruti Sakhare, Dipti Sakhare

A Robust Approach of Estimating Voice Disorder Due to Thyroid Disease

Thyroid is butterfly-shaped gland present in the lower anterior of the neck. The main root of thyroid disease is the improper working of thyroid gland. Thyroid disease is mainly categorized into two types, i.e., hypothyroidism and hyperthyroidism. In this study, voice samples for two disorders—hypo and hyper along with normal voice samples are considered. A databank is created for three classes—normal, hypo, and hyper. The structure of the robust approach of diagnosing thyroid disease contains four stages. In the first stage, preprocessing is performed by considering framing, windowing, and filtering. In the second stage, feature extraction is performed by using mel-frequency cepstral coefficient (MFCC) method. In the third stage, classification is achieved by using combined classifier, i.e., support vector machine (SVM), and hidden Markov model (HMM). In the fourth stage, performance evaluation for diagnosing thyroid disease is achieved by estimating accuracy, confusion matrix, and precision. The classification accuracy of a robust approach for diagnosing thyroid disease is obtained about 97.28%.

Namrata V. Kanase, Satyajit A. Pangoankar, Ashish R. Panat

Smart Glasses: Digital Assistance in Industry

New media developments have revolutionized the behavior of people in an unprecedented technique in the latest decades. Mobile phones have created an always online mentality. However, what is next? Recent developments underline the increase of a technology known as “Wearable devices.” Augmented reality smart glasses (ARSG) such as Microsoft HoloLens and Google Glass are very good examples of these technology. It provides huge potential for innovation for firms and manufacturing industries. ARSG is becoming very common and important technology that promotes shop floor operators to fulfill industry 4.0 requirements. Augmented reality is currently an interesting and hot research topic in manufacturing industries. The main goal of this research paper is to improve the use of smart glasses for operator training with augmented reality. It will assist to increase effectiveness and shorter learning times for the individual operator. ARSG products available in the market are very expensive. It will help to find an affordable solution for the industries. It provides new methods for reducing the efforts of the operators working online. It mainly focuses on minimizing disadvantages of the existing products.

Trupti Sutar, Savita Pawar

Implementation of Hand Gesture Recognition System to Aid Deaf-Dumb People

In recent years, the population of deaf-dumb victims has increased because of birth defects and other issues. Since a deaf and mute person cannot talk with an ordinary person in order that they ought to rely on some kind of communication system. The gesture shows some physical movements of the hand that convey a piece of information. Gesture recognition is the analytical interpretation of the movement of an individual through an information processing system. Linguistic communication provides the most effective conversation platform for the mute person to speak with an ordinary person. The aim of this paper is to build up a time system for hand gesture recognition that acknowledges hand gestures and then converts them into text and voice. In this paper, efforts have been done to detect 8 different gestures. Each gesture has assigned unique sound and text output. In experimental results, 800 samples were taken into the consideration out of which 760 samples were detected correctly and 40 samples were detected wrongly. Hence, the proposed system gives accuracy of 95%.

Supriya Ghule, Mrunalini Chavaan

Robust Underwater Animal Detection Adopting CNN with LSTM

Underwater detection of objects valuable problem for many civil and military applications such as hydrographic surveys for the purpose of ensuring navigation. The objective of this work is to ensure flexibility, speed, and precise recognition of object underwater system for use in a variety of low-level underwater images captured. These imaging systems are used in separate occasions and under distinct weather and bathymetric circumstances from underwater imaging systems. This paper discusses the use of deep learning in the latest past to analyze underwater imagery. The methods to analyze are classified according to the object of detection, highlighting the characteristics and architectures used for deep learning. In the evaluation of digital sea bed imagery using deep neural networks, it is found that there is excellent scope for automation, particularly for the detection and tracking of detected object is harmful or non-harmful.

Harishchandra Jagtap, Mrunalini Chavaan

Face Recognition Using Golden Ratio for Door Access Control System

The paper introduces a method-based correspondence scheme that integrates a permutation of Viola–Jones face detection method with characteristics of extracting golden ratio. The purpose of this article is to help users improve the security of sensitive places through facial recognition. We come to resolve the issue of low precision in the suggested technique. Here, we propose a new scheme for aligning the face using Viola–Jones and support vector machine face detection method followed by extraction technique of golden ratio function. The suggested technique is very effective, more realistic and accurate compared to other face detection techniques. The module contains a secure face identifier. The scheme is designed to satisfy the requirements of classification face to face in real scenario.

Prajakta S. Gaikwad, Vinayak B. Kulkarni

Efficient Design of Drone Flight Control Using Delay Tolerant Algorithm

As the activities of drones which are specifically concerns to flying, racing are growing, drone crashes are also growing. Many of these crashes are linked with operational problems, and hence, improving drone controls is urgently needed. The main objective of this entire study was to reduce the accident of drone due to mistakes made by a human. The approach given in the paper describes a composable pipe model for task scheduling. The primary objective of this system is to enhance the efficiency of task scheduling; for this, we have used two proportional integration derivative (PID), controllers. Two end-to-end terms are analyzed using the pipe model: reaction time and freshness time. We have used Cleanflight control firmware with Real-Time Operating System (RTOS). The experimental results convey that the latency time and delay time of task are getting reduced.

Priyanka Dange, Bhairavi Savant

Adaptive Background Subtraction Models for Shot Detection

The paper projected a unique prospective on background subtraction for object identification which is in motion as a structural block for many multimedia application beaning the primary applicable stage for successive detection, classification and analysis of actively of task. Since color information is not sufficient for addressing issue like sadden changes in elimination or visibilities in foreground object and color conflict. We have projected this work in which subtraction of background for detection of shot boundary in video is based on adaptive technique. This method depends on detecting the difference of mean gray value of current frame and previous frame is incorporated the result as arranging. We have calculated the result for five video inputs and evaluated outcomes interims of precision, recall and F1 measure.

Dattatraya A. Jadhav, Yogeshkumar Sharma, Parul S. Arora

Automatic Gear Sorting Using Wireless PLC Based on Computer Vision

Gears are the most important components of the device and are usually used in the design of transmission of cars and other pivoting devices. In this paper, computer vision is suggested to sort out the defective equipment based on image processing depending on their amount of teeth image processing instrument and sensory circuitry used to solve the gear sorting issue. Through less human participation, a programmable logic controller (PLC) is used in sectors to automatically execute the entire manufacturing cycle to prevent human errors. A low-cost automation (LCA) has emerged in sophisticated technology that is used to prevent wiring composition and to effectively acquire control over the process. This paper involves converting wired PLC into wireless PLC by interfacing the PLC with the Wi-Fi module. To enable real-time surveillance and control of the system of equipment sorting via Wi-Fi module interfacing with PLC.

Yogesh Darekar, Smita Kulkarni

Statistical Validity of Presmoking and Postsmoking Impact on Heart Rate Variability Among Middle Age Men

Background: Cigarette smoking is associated with various forms of an acute heart attack such as myocardial infarction, arrhythmia, and atrial and ventricular fibrillation. Increased sympathetic activity triggered by cigarette smoking is one of the major risk factors for cardiovascular extinction. Objective: To analyze the acute effects of smoking on the control of neuro-cardiovascular by evaluating the time domain, frequency domain, and nonlinear HRV indices in middle-aged smokers. Method: Thirty-six male participants of age between 40 and 60 were evaluated and divided into two groups, i.e., control and smokers. The ECG recorded for 15 min from control and smokers male participants. In the case of smokers, data recorded before smoking and 10 min after smoking. The heart rate variability (HRV) indices in the time domain (mean HR, mean RR, SDNN, and RM SSD), frequency domain (LF, HF, and LF/HF ratio), and nonlinear parameter (SD1, SD2, SD1/SD2, Poincare plot, detrended fractal analysis, approximate entropy, and sample entropy) were evaluated. The electrocardiogram (ECG) was recorded using our own designed ECG module, we used lead II data for experimentation. Statistical analysis performed between control versus presmoking using independent student ‘t’ test and presmoking versus postsmoking using dependent student ‘t’ test. The presmoking values are considered as baseline values. Result: When control group compared with the smoking group SBP, DBP, mean HR, LF (ms2), LF(nu), and LF/HF ratio were significantly increased and also mean RR, SDNN, RM SSD, TP (ms2), HF (ms2), HF (nu), SD1, SD1/SD2, AppEN, and SampEN significantly decreased after smoking. Conclusion: The results of this study show that smoking has an acute effect on the autonomy of the brain, causing impaired vagal activity and an overbalance of sympathetic function.

S. R. Rathod, C. Y. Patil

Transition Based Odd/Full Invert Coding Scheme for Crosstalk Avoidance and Low Power Consumption in NoC Links

In this paper, a Transition Based Odd/Full Invert (TBO/FI) coding scheme, which focuses on crosstalk avoidance and low dynamic power consumption in NoC links is proposed. This scheme is designed and implemented at both architectural and logic level and is evaluated using synthetic traffic scenarios for both 4- and 8-wire links. All the evaluations are performed for the worst-case, the best-case, and the typical-case scenarios. TBO/FI coding scheme has the maximum reduction percentage for both the switching activities in all the cases for 4-wire link and same is true for 8-wire link except for the worst-case scenario. TBO/FI coding scheme allows NoC power savings of up to 25.5% and 40.4% for 4- and 8-wire links with worst-case scenario and with other scenarios, NoC power consumption increases. However, this increase is lower than that of other existing schemes. These results are achieved despite the NoC router area and power overheads of 117% and 50% for 4-wire and 52% and 26% for 8-wire link, respectively. NoC router area and power reduces by 16% and 30%, respectively, with increase in link width and this reduction is more compared to all other schemes.

M. Vinodhini, N. S. Murty

Feature-Based Model for Landslide Prediction Using Remote Sensing and Digital Elevation Data

This study aims to generate landslide susceptible maps and landslide hazard zonation maps using the digital elevation model for the prediction of future landslides. The landslide zone is based on the qualitative and quantitative factors combined using the weighted sum of the different features and hydrological parameters. The main aim of the research is to discover the damaged areas with the help of detailed field observation of prior and post landslide events. Landslide hazard zonation is a map classified into six different zones ranging from very low hazard zone to scars hazard zone and to represents the prediction of future landslide occurrence under the area of the study. The result of this study shows that a very high and scars susceptible region depicts a higher chance of landslides.

Litesh Bopche, Priti P. Rege

Acoustic Classification of Bird Species

The main focus of this paper is identification of bird species or even individual birds on the basis of their sounds. This work compares an audio signal of an unknown bird to a database of known birds. The system has two modes of operation: training mode, and recognition mode. In the training mode, ate a feature model of the available bird sounds in the database is created. The recognition mode will use the information obtained from the training mode to isolate and identify the bird. Mel frequency Cepstral Coefficients and Gammatone frequency Cepstral Coefficients have been employed as feature sets for classification. The classification accuracies are evaluated using Support Vector Machine and Artificial Neural Networks.

Rashmika Patole, Priti Rege

Analysis of Chronic Joint Pain Using Soft Computing Techniques

In recent years, chronic pain in joints is the most common type of musculoskeletal pain among office workers. As the work culture in the office has been changed, it is the common demand that everyone has to work with a computer. In office, the worker has to sit continuously on a personal computer at least for 3–4 h. Most of the time, this leads to chronic pain in joints, generally after the period of few years of repeated exposure to such kind of regular engagement of user-computer interfacing, which is consequently converted into a serious health concern in the considerably large-sized office community. In this paper, an attempt has been made to provide a tool to assess the level of pain in the incumbent worker to have a reliable measure and at the disposal of medical practitioners to know the degree of pain to diagnose the severity of the health concern attached to it and have an optimum analysis which is not available even today. Around 250 samples have been taken and these cases have been categorized based on level of severity of pain, i.e. normal, medium, and severe. The results obtained show a considerable difference between mean and standard deviation values of pain level. Based on the data analysis of rectified EMG data, it has been observed that there is a notable difference in the Power Spectral Density of clean EMG signal w.r.t mean and Standard Deviation. The accuracy of the experimentally determined pain intensity level is a more reliable source for medical practitioners to make their diagnosis more objective one and this can be used as one of the measures to calibrate the pain-intensity.

Shailaja Suresh Patil, Shubhangi B. Patil

Emotion Recognition using Gamma Correction Technique Applied to HOG and LBP Features

Human social interaction, especially facial expressions, is often influenced by non-verbal communication. The surrounding people often watch the face in day-to-day group interaction to understand the inner feelings of a person. Face thus forms an essential source of human emotion recognition that is generally categorized as a surprise, fear, anger, disgust, sad, and happy. Recognition of emotions plays an important role in a variety of fields in behavioral science. In this paper, median filtering is used for pre-processing of an input image. Watershed segmentation is used before extracting features to obtain the necessary image properties. Gamma correction is implemented in this paper, and features are obtained, including Histogram of Oriented Gradient (HOG) and Local Binary Pattern (LBP) techniques. The performance of LBP and HOG is evaluated. kNN and SVM are used as classifiers for comparing the efficiency of recognition. The overall accuracy, along with precision, recall, and f-score has been computed and compared. It is found that all performance parameters with gamma correction give better performance compared to without gamma correction.

Vishal D. Bharate, Devendra S. Chaudhari, Mayur D. Chaudhari

Scalable and Rapid Fault Detection of Memories Using MBIST and Signature Analysis

A novice scalable MBIT unit with MISR signature analysis is proposed in this project. Enhancement of MBIST architecture with signature analysis significantly improvises resolution of fault detection in memories in comparison to conventional MBIST Architecture. The proposed MBIST Algorithm is optimized by using only 14 states for 7 March algorithms, hence improvising the scalability of the MBIST without area overhead. The proposed architecture has an interface check MISR which features isolated detection of fault in memory interface and memory, hence improving resolution and accuracy. Rapid Fault check is enabled by Memory pre-check using MISR. Benchmarking for this project in terms on the accuracy, feature enhancement, scalability and the ability to detect early faults in less time are done via Random constrained verification of the developed RTL Model....

Midhun Sasikumar, Ramesh Bhakthavatchalu, K. N. Sreehari, Arjun S. Kumar

Analyzing Vocal Tract Parameters of Speech

Speech sounds produced by human depends on movement of various articulators. Dimensions and shape of the various elements of speech production organs also have impact on nature of speech produced. Vocal tract plays a major role like characteristics of linear time invariant system. In this paper, analysis of vocal tract parameters, in terms of resonances of vocal tract, also referred as formants is done. Of the different speech sounds, vowels carry most significant clues of speech. Despite specific properties of different vowels, there is much variability of vowel characteristics among speakers. It gives characteristics of vocal tract related to acoustic resonances. At large, first four formants are useful to categorize vowel sounds. The variation in formants for same speech sounds is a challenge to speech recognition algorithms in which vowel spectral characteristics are assumed to be invariant among speakers. But the same variability of formants among speaker is useful in speaker recognition. The method used for formant estimation is based on designing all-zero filters to track the formants and voicing detection-based formant extraction filters to estimate the first four formants. Vocal tract parameters in terms of formants are analyzed using twelve vowel sounds from different speakers. From the experimental analysis, it is observed that the first formant specifically represents the pertinent characteristics of vowel speech, whereas there is very little consistency of higher-order formants for same speech sounds by different speakers, indicating the impact of physiological nature as well as behavioral aspects of individual on nature of speech produced.

Effect on Quality of 3D Model of Plant with Change in Number and Resolution of Images Used: An Investigation

3D reconstruction of plants is a challenging task. It is essential in non-destructive plant growth monitoring systems and important to reconstruct plant features precisely. Two parameters which critically affect quality of the 3D model are change in the number of input images and the resolution of the images. In this paper, the effect on 3D models of these parameters is analysed. This analysis enables an appropriate number of input images and resolution of the images to be determined for a precise 3D model. To validate the results, the plant stem height and number of leaves extracted from reconstructed 3D models were compared with ground truth values. We used descriptive statistics for validation and achieved high correlation between extracted and ground truth values.

Abhipray Paturkar, Gourab Sen Gupta, Donald Bailey

Comparative Analysis of Least Squares Method and Extended Kalman Filter for Position Estimation in GPS Receiver

GPS is a system of obtaining the position of any object on or above the earth surface. Global Positioning System (GPS) has been incorporated into many devices in everyday life. However, GPS receiver design is challenging depending on the user’s operating environment. The accuracy of GPS position estimate is affected by various factors like ionospheric delay, tropospheric delay, various multi-path effects, and number of satellites in view and navigational solution employed. Multipath propagation to low signal strength are examples of some of these challenges. This paper provides a comparative analysis of position estimation techniques in a GPS receiver. These techniques are the Least-Squares (LS) method and Extended Kalman Filter Method (EKF). In this, the data possessed with a dual-frequency GPS receiver is placed at the reference point (X-1687535, Y-5809975, Z-2014102). For this reference point iterative assessments of satellite transmission moment, receiver time, and position are carried out to determine instantaneous estimates of the receiver location. The work explains the design and implementation of a software-defined GPS receiver in real-time. We use five satellites to estimate the position of the receiver. The performance evaluation of position estimation accuracy over the region is carried out based on position coordinates, root mean square error (RMSE), and standard deviation. The experimental evaluation demonstrates that the Extended Kalman Filter provides a more precise and stable estimate than LS method.

Jyoti S. Kavathekar, Ashwini M. Deshpande

Fair Scheduling Non-orthogonal Random Access for 5G Networks

The potential growth of communication medium and its associated devices are rapidly increasing day by day. A most powerful communication support with 5G network features is required nowadays to resolve all communication overheads. The traditional method of embryonic data transfer between base station and the user device follows channel allocation strategy, which is internally used for the maintenance of random entry of data packets resulted in poor scalable nature, poor traffic handling, latency, and so on. The efficiency of network technologies with 5G support suffers from major issues such as the network performance, collision, and delay that are needed to be concerned with. To combat this, the technique of decoding of two or more packets is followed; hence, the concept of NORA was proposed. In analysis of the existing NORA method, Zadoff–Chu sequence was used to obtain the performance of NORA in 5G which gave better performance yet, resulting in collisions. Keeping these issues in consideration, an efficient communication model called fair scheduling non-orthogonal random access (FSNORA) is proposed. In the proposed method, we have scrutinized NORA to improve the fairness of scheduling by using round-robin algorithm so as to make a contentionless system and improve the channel enhancement of 5G networks. Simulation results convey that there is an improvement in the access delay, success probability, throughput, collision probability, and the preamble transmissions in context with contention compared to the existing NORA method.

Mansi N. Purohit, T. K. Ramesh

Analysis of Tree-Based Classifiers for Web Attack Detection

In the last few decades, the increase in the use of Web services has led to an increase in the threats of Web attacks. The severity of such Web attacks is increasing day by day. Intrusion detection systems play a crucial role in identifying Web attacks in proactive manner. There are large numbers of features present in the network traffic. Identification of relevant and irrelevant features is crucial task in machine learning. This paper proposes a Web attack detection system that consists of preprocessing, feature selection, reduced dataset, and tree-based classifiers. The system uses information gain filter method to select relevant features for the classification of Web attack. The system is tested on CIC-IDS-2017 dataset. The experimentation results show that random forest produces high precision of 74.5% for brute force, and J48 produces high precision of 63.8% and 87.5% for cross-side scripting (XSS) and SQL injection (SQLi), respectively, with 65 selected features.

Deshmukh Surbhi, Kshirsagar Deepak

Implementation of Random Pulse Width Modulation Techniques for the Open-End Winding Five-Phase Motor Drives to Reduce Acoustic Noise and Harmonic Distortion

Multi-phase machine drives nowadays found its applications in electric vehicles, ship propulsion, space craft, drilling machines. High-speed switching of the space vector pulse width modulation techniques (SVPWM) generates high rate of change of currents and voltages in the inverters, which causes serious electromagnetic interference. These generate harmonics in output voltages and currents and high acoustic noise is produced. All these applications produce high acoustic noise and high harmonics distortion which disturbs the human being and some electronics devices that are nearer to the drive. Hence, this paper provides a solution by implementing random pulse width modulation (PWM) techniques for the dual-inverter fed open-end winding (OeW) five-phase induction motor drives to reduce acoustic noise levels that are produced in output effective phase voltages and currents. Randomization in modulating signal; randomization of high-frequency triangular carrier signals; both modulating signal and carrier randomization; variable switching frequency randomization of both modulating signal and high-frequency triangular carrier signals methods are implemented. To show the effectiveness of the proposed random techniques over the conventional continuous PWM techniques, simulation and experimental studies on 1 Hp proto type five-phase motor are presented.

J. Balakrishna, Teegala Bramhananda Reddy, Marapu Vijaya Kumar

A Frequency Reconfigurable Antenna for Sub-GHz and TV White Space Applications

TV White Space (TVWS) band of frequencies are those rendered surplus after conversion of analog TV transmission to digital. The proposed antenna is a meander line antenna (MLA), with a defect ground structure designed for TVWS band viz 470–890 MHz. The antenna has dimensions of 108 × 44 mm2 printed on FR-4 epoxy substrate with dielectric constant of εr = 4.4 achieving size reduction upto 35%. Frequency is switched electronically, using four RF-PIN diodes, mounted in the slots on the ground plane. The simulated resonant frequencies obtained for the different PIN diode biasing combinations are 0.840, 0.660, 0.8560, 0.6460, 0.8700, and 0.828 GHz. The simulated return loss values are well below −10 dB. The measured results of the printed antenna closely follow the simulated ones.

Sanjeev Kumar, Rohit Khandekar, Neela Rayavarapu

Human Activity Recognition Using Positioning Sensor and Deep Learning Technique

In this paper, Long Short Term Memory (LSTM) deep learning model is used to identify human activities using sequential data obtained from cameras, wearable sensors, or other modalities. The proposed method recognizes human activities by optimizing hyper-parameters for the chosen deep learning model. The proposed approach is validated using public domain UTD MHAD dataset and found to be more accurate outperforming the state-of-the-art in terms of accuracy of activity recognition. Datasets categorized as the depth, skeleton, and inertial modalities have been analyzed for all available features. The dependency of the proposed deep learning model on the hyper-parameters is investigated extensively and discussed in detail. Experimental results depicted in this paper demonstrate the fact that the proposed method is a better choice for indoor activity recognition.

Aarati Mohite, Priti Rege, Debashish Chakravarty

A Discriminative Model for Multiple People Detection

Group activity recognition is becoming more important day by day for video surveillance, sports analytics, etc. Monitoring various cameras manually through human resources is a complex job; due to this, computer vision algorithms are being developed to perform lower and higher level tasks. This paper presents multiple people detection using the histograms of oriented gradients (HOG) feature descriptor algorithm through a support vector machine (SVM) based on the different group action class of persons. Multiple people detection for group action identification is a complex problem as accurate detection of individual persons requires extensive computation. To achieve multiple people detection for group activity, HOG feature extraction is proposed. HOG is precise and accurate person detection algorithm in the recent computer vision application. In addition, thresholding algorithm is implemented to collect the HOG feature vectors of definitely detected windows and firms the pathway followed by a person in the video frames. The proposed algorithm is evaluated through different aspects like group action categories and existence of occlusion over Haar and HOG features.

Jal Sanchay—A Novel Approach for Water Usage Monitoring

Water is one of the, most important resources on earth. With rapid increase in the world population, water consumption is increasing drastically. People now a days always want something new that can make their life easier. The technological advancements of embedded system as well as articulation of communication by sensing techniques are taking a huge role in recent days. LabVIEW which is a system engineering software for applications requiring test measurement and control with rapid access to hardware and data insights. Design of various virtual instruments (VIs) in LabVIEW provides a strong graphical tool and a platform where automated water usage monitoring system can acquire efficiently as well as with accuracy. In this proposed system GPRS enabled sensors are used to sense water flow in every outlet. The server continuously monitors and collects the data over the internet and tracks usage of water at every outlet via a wireless sensor node. At the point when water is utilized at overabundance it is shown and an alarm is sent to the user. The user can persistently monitor the water usage and the wastage through their mobile devices.

An Algorithm for Skew Angle Estimation and It’s Application Domain

Angle detection and estimation is important in various fields like line following mobile robots, document analysis, construction sites etc. There are several methods and algorithms proposed earlier to do that. This paper represents a new algorithm to estimate skew angle and its possible applications in different fields. In this algorithm, angle in the preprocessed black-white image is estimated using matrix traversal and slope equation from Euclidean geometry. This algorithm can measure the angle between −90 $$^{\circ }$$ ∘ and +90 $$^{\circ }$$ ∘ efficiently.

Unnati Raju Kulkarni, Hemant Goraksh Ghuge, Revati Anand Kulkarni, Kirti Vasant Thakur

Analysis of Accuracy of Supervised Machine Learning Algorithms in Detecting Denial of Service Attacks

Intrusion Detection Systems are considered to be one of the primary methods for security attack detection. It is very challenging to design and implement intrusion detection systems that can detect the newer variants of security attacks with greater accuracy. This paper focuses on the detection of the Denial of Service Attacks in particular by DoS attack tools like Goldeneye, Slow HTTP test, Slow Loris, and Hulk. Further, this paper also focuses on the detection of one of the most important Web Application security attack due to the Heartbleed vulnerability. We have used the supervised machine learning algorithms like Support Vector Machine, Decision Tree, K-Nearest Neighbor for analyzing the accuracy of these models in classifying multi-class problems. One of the highlights of this paper is that the CICIDS2017 attack dataset has been used for evaluating the accuracy of various classification models. This research work holds significance as this focuses on the classification of attacks into five categories rather than a binary classification problem which is the focus of majority of the research works.

Deepa Krishnan

An Improved Carrier Frequency Offset Estimation Under Narrowband Interference in OFDM Cognitive Radio

In recent years, overcrowded unlicensed spectrum is devastating spectral efficiency of communications in regional and rural broadband wireless networks. Cognitive radio allows opportunistic use of a licensed spectrum without interfering with primary users (PU) which overcome the scarcity problem of the available spectrum. The occurrence of carrier frequency offset (CFO) degrades the performance of the orthogonal frequency division multiplexing (OFDM). OFDM fulfills the requirements of cognitive radio, and hence OFDM is an appropriate choice for cognitive radio. When OFDM is used for cognitive radio applications, sensitivity to frequency offset remains an issue. This paper surveys various techniques present to estimate carrier frequency offset for OFDM cognitive radio. It covers required parameters to estimate, i.e., training symbols, estimation range, and complexity. This paper also presents simulation results that show extended estimation range of the frequency offset at good performance in the presence of narrowband interference. This method uses correlation among L identical parts of the training symbol at the receiver side to estimate the frequency offset. The estimation range is achieved up to ± L/2.

Digital Image Watermarking by Fusion of Wavelet and Curvelet Transform

A new hybrid watermarking algorithm using discrete wavelet transform (DWT) and curvelet transform (CT) is implemented in this paper. Curvelet transform is developed to be more efficient than traditional transformations to represent edges along curves. In the proposed work, image watermarking is achieved by applying hybrid transformation and then evaluating the efficacy of the method. Initially, image is decomposed using DWT and then fast curvelet transform is applied on cD (diagonal edges details) sub-band. Secondly, lower coefficients are selected to embed binary-encoded watermark bits. Efficiency of algorithm is tested with different texture images and by varying size of watermark. The effectiveness of proposed algorithm is evaluated by applying attacks like resize, median filtering, and addition of noise. The results shows that, it is sustaining over different texture images even with increase in watermark embedding capacity. The proposed method results in 34.33 dB peak signal-to-noise ratio (PSNR) for high texture image, 47.77 dB for medium texture image, and 48.62 dB for low texture image over maximum size of watermark. It has good robustness against attacks and on large embedding capacity.

Machine Learning Feature Selection in Archery Performance

Successful sports performance depends on several physiological and physical fitness components. It is essential to know which fitness features are most important for performance. Sports fitness components are often multicollinear, and the relationship is complex. So there is a need to use more sophisticated methods that can deal with complex multicollinear data. Hence machine learning algorithms are used along with conventional statistical methods for important physiological and physical fitness feature selection in the archery performance of Indian archers. Recursive feature elimination and Boruta algorithm using random forest along with conventional statistical methods are used for feature selection. The root mean square error of stepwise regression was 21.262, recursive feature elimination with 15 features was 19.450 and random forest with 15 features was 8.417 in the training dataset. Further, the root mean square error for random forest with eight confirmed important features was 9.003 and 8.716 for ten non-rejected features in the training dataset. Out of fifteen features, eight features confirmed important are maximum bow hold time, sub-maximal oxygen intake, peak power, average power, core abdominal strength, age, weight, and body fat, while acceleration speed and maximum oxygen intake are tentatively important. Machine learning Boruta algorithm using random forest performs better than traditional statistical and recursive feature elimination method for selecting features as well as predicting performance in unseen data. Thus, eight important features identified through Boruta algorithm are useful to develop battery of test, monitor athletes, and alter the training regimens in real-time and talent selection in the archery.

Uma Mahajan, Anup Krishnan, Vineet Malhotra, Deep Sharma, Sharad Gore

Skin Lesion Classification Using Deep Learning

Skin cancer is a common disease and considered to be one of the most prevalent forms of cancer found in humans. Over the years various imaging techniques have shown improvement and reliability in diagnosis process of Skin Cancer. However, quite a few challenges are being faced in generating reliable and well-timed results as adoption of clinical computer aided systems is still limited. With the recent emergence of learning algorithms and its application in computer vision suggests a need for combination of sufficient clinical expertise and systems to achieve better results. Here we attempt to bridge the gap by mining collective knowledge contained in current Deep Learning Techniques to discover underlying principles for designing a neural network for skin disease classification. The solution is based upon merging of top-N performing models used as a feature extractor and a SVM to facilitate classification of diseases. Final model gave 86% accuracy on ISIC 2019 dataset along with high precision and recall values of 0.8 and 0.6, respectively.

Vehicle-to-Vehicle Driver Safety-Related Data Transmission and Reception Using Li-Fi Technology

Light Fidelity Technology is also known as Visible light communication is the form of wireless communication which uses visible light to transfer information such as digital data, Audio and video as well. Light is modulated and amplified to attain desired speed and distance. Vehicle-to-vehicle communication is the technology in which one vehicle transmits and receives data to and from other vehicle so that they share data between each other and will be able to assist each other. Proposed system uses Li-Fi for vehicle-to-vehicle communication system uses Li-Fi module which can be mounted in the Headlamp as well as tale-lamps of the Four-wheeler which will help to transmit real-time information such as speed data, anti-lock braking data, Turn indication, certain emergencies in car, tire related data. System is made for such small applications which can be directly implemented in car with little modifications in Hardware and Software.

Snehal Pacharne, Vinayak Kulkarni

A Novel Approach for CBIR Using Four-Layered Learning

Content-based image retrieval (CBIR) comprises recovering the most outwardly comparative images to a given question image from a database of images. CBIR from therapeutic image databases does not plan to supplant the doctor by anticipating the sickness of a specific case however to help him/her in analysis. The visual attributes of an ailment convey analytic data, and periodically outwardly comparative images relate to a similar infection class. By counseling the yield of a CBIR framework, the doctor can acquire trust in his/her choice or considerably think about different potential outcomes. With high-dimensional information in which every point of view on information is of high spatiality, determination of highlights is imperative to further build the aftereffects of bunching and characterization. To ease the enthusiastic miscellany in the precision of image retrieval, we developed another graph-based learning strategy technique to successfully recover images from remote detecting. The proposed strategy utilizes a four-layered framework that joins the feature level fusion of Gabor and ripplet Transform of selected query along with SVR. In the first layer two image sets are retrieved utilizing the Gabor and Ripplet-based wavlet Decomposition separately, and the besides, the top ranked retrieved images from both the top up are further used to find their queries. Using each individual part, the chart grapples recoup six image sets from the image database as an augmentation request in the subsequent layer. The photos in the six image sets are evaluated for positive and negative data age in the third layer, and Simple MKL is associated with gain proficiency with the proper inquiry subordinate combination loads to accomplish the last consequence of image recuperation. This research is based on building fully-automatic four layer systems capable of performing large-scale image search based on texture information. An effective four layer architecture with the application of SVR was proposed in this study for the purpose of retrieving images from CIFAR dataset.

Shweta Salunkhe, S. P. Gaikwad, S. R. Gengaje

Design of a Power Efficient Multiband Patch Antenna

A psi-shaped antenna design is proposed using Ansoft’s HFSS software in this paper. In this structure, there is a rectangular slot above the monopole antenna and two metal strips besides that structure on the substrate with a slotted rectangular defected ground structure for wireless application. This monopole antenna can be used for WLAN, WiMAX applications. The structure is designed and optimized to operate at 2.5, 3.4, and 5.6 GHz frequencies. The size of the antenna is 34 × 18 × 1.6 mm3. The presented multiband antenna has been designed, simulated by using HFSS software. The antenna has isolation more than 20 dB and peak gain is 3.81dBi. The antenna utilizes microstrip feed. The tri-band good resonance is obtained by rectangular strips. The performance parameters are satisfying the requirements. The gain, return loss, radiation pattern, efficiency, VSWR, 3D polar plot results have been studied through HFSS software. The simulation results satisfy general requirements for commercial use.

Brain Activity Analysis for Stress Recognition

The stress is the major problem that occurs in daily life, which affects on the physical and mental health. There are various methods for detection of stress. In this paper, EEG signal analysis is used for stress recognition. The output of a neurosky mindwave mobile 2 sensor is waves like alpha, beta, and gamma in a specific range. By analyzing these values and keeping a threshold, the dataset formation occurs, and further to train the data, artificial neural network technique (RBFN algorithm) is used. The system learns and is trained using RBFN. The states (stress) are detected. The work is tested on hundred cases and found 80% accurate.

Aishwarya Wakale, Usha Verma

Deep Learning-Based Paperless Attendance Monitoring System

This paper concerns about the paperless attendance monitoring system. Major focus of this system is on the concept of face recognition. The goal is to mark the attendance without disturbing the class by using face recognition technique. Marked attendance using this monitoring system can be directly sent to the mentor/coordinator through email. Face recognition uses biometric features, and it extracts person’s facial features and stores data. Faces will be recognized based on deep learning algorithm to train the system and to compare or test to identify the person. Time is an important factor in recognition. This factor has been brought up to get the attendance in hand of the respected person instantly. To implement this efficiently, we have gone for deep learning model based on CNN. One important aspect is to deal with the results of the live detection and the immediate prediction of the person. The proposed system can identify an individual from various angles and positions. One of the important parameters that decides the accuracy is the quality of image, better the quality better the accuracy. Quality of the captured image defines the accuracy of the system. After implementing CNN on created database, we achieved an accuracy of 80.5%. Accuracy value can change depending on the testing database.

Pallavi Derkar, Jitesh Jha, Mayuresh Mohite, Rushikesh Borse

Image Analytics to Detect Cigarette in an Image Using Deep Learning

Significant number of modern films depict some form of tobacco use, but rarely depict its real-life consequences such as addiction, illness and death. As per [1], anti-tobacco health warnings are mandatory for scenes depicting smoking scenes. In this paper, an automated recognition system is proposed to identify images with smoking activities and tag them accordingly. The proposed approach implements the technique of object detection based on deep learning. Convolutional neural network is used to generate feature maps from the images. These machine-learnt features are used to classify the images. The system can detect the smoking events of uncertain actions with various cigarette sizes, colors and shapes. We have experimented our work by applying the proposed approach to two real-world datasets and that have demonstrated the effectiveness of our solution with a decent model accuracy.

Abhijeet Kharade, Kumar Abhishek, Debaraj Dwibedi, Siddharth Mehta, Hemanth Meruga, Pratap Gangula, D. Narayana, Rushikesh Borse
Weitere Informationen