Abstract: Diabetic retinopathy is when damage occurs to the retina due to diabetes, which affects up to 80 percent of all patients who have had diabetes for 10 years or more. The expertise and equipment required are often lacking in areas where diabetic retinopathy detection is most needed. Most of the work in the field of diabetic retinopathy has been based on disease detection or manual extraction of features, but this paper aims at automatic diagnosis of the disease into its different stages using deep learning. This paper presents the design and implementation of GPU accelerated deep convolutional neural networks to automatically diagnose and thereby classify high-resolution retinal images into 5 stages of the disease based on severity. The single model accuracy of the convolutional neural networks presented in this paper is 0.386 on a quadratic weighted kappa metric and ensembling of three such similar models resulted in a score of 0.3996.

Abstract: Skin diseases are most common form of infections occurring in people of all ages. As the costs of dermatologists to monitor every patient is very high, there is a need for a computerized system to evaluate patient's risk of skin disease using images of their skin lesions. Many researchers have used different preprocessing, segmentation and classification techniques to determine whether a skin image suffers from diseases or not. Feature extraction is very important for predictive modeling applications. Feature extraction in image processing is a method of capturing visual content of images for indexing and retrieval. Primitive image features can be either general features, such as extraction of color, texture and shape or domain specific features. Texture based features are widely used in image analysis for medical diagnosis. This paper presents a comprehensive survey of texture based feature extraction for detection of skin diseases and proposes a system based on the findings.

Abstract: Internet of Things (IoT) is growing as an attractive system paradigm. There is a lot of hype around the internet of things (IoT) and it continues to evolve as we move beyond humans talking to machines. IoT has interconnections through the physical, cyber and social spaces. Things used in IoT are sensors and actuators, mechanical devices and networking includes gateways, wireless infrastructure. Most of devices among them are resource constrained. During the interaction between devices, IoT gets suffered from severe security challenges. Complicated network produces potential vulnerabilities referred to heterogeneous devices, sensors and backend systems. So to realize the dream of internet of things secured device to device communication is expected. Security of resource constrained networks becomes prime important. Many existing mechanisms gives security and protection to networks and systems but they are unable to give fine grain access control. In this work, we focused on CoAP based framework to give service level access control on resource constrained devices. It gives fine grain access control on a per service basis. ECDSA is used to improve privacy of the system. Performance of CoAP based framework is compared and analyzed with existing security solutions. Test results are presented which shows that communication overhead and authentication delay are less than the existing system. Hence security performance of system gets improved. The goal is to present comprehensive security framework for low power networks consist of resource constrained devices.

Abstract: In CPU scheduling various algorithms exist like FCFS (First come first serve), SJF (Shortest job first), SRTF (Shortest remaining time first), Priority Scheduling, Round Robin (RR), MLQ (Multilevel queue), MLFQ (Multilevel feedback queue) scheduling. Multilevel Feedback Queue (MLFQ) algorithm allows the switching of processes between queues depending on their burst time. The processes switch to the next queue when burst time is greater than time quantum. Each queue can define its own scheduling policy. In this paper we have implemented MLFQ technique using small burst time for the first queue thus making it analogous to RR scheduling and using SJF prior to RR from second queue onwards gives better CPU utilization. Dynamic time quantum is also used which further improves the efficiency of the scheduling. Here the dynamic time quantum of the queues is calculated based on the burst time of the processes. Time quantum of the second queue is the burst time of the (2n/3)th process (where n is the number of processes remaining after the execution in the first queue) and time quantum of the third queue is burst time of the largest remaining process. Thus 66% of the processes get executed in the second queue and remaining processes in the last queue thus preventing the problem of starvation of huge burst time processes.

Abstract: In multitasking operating system, processes don't run simultaneously, but switch very expeditiously. Thus, all the processes share the CPU time. It is the job of the scheduler to select a process from the ready queue and place it into the memory based a particular strategy known as Scheduling Algorithm. There exist many Scheduling Algorithms such as First Come First Serve (FCFS), Shortest Job First (SJF), Round Robin (RR), Priority Scheduling, Multilevel Queue Scheduling (MLQ). This paper proposes a new strategy consolidating three of the existing scheduling algorithms to create a new strategy that reduces the time a process spends in waiting state. The proposed algorithm also reduces the number of context switches to provide a fair, efficient and methodical scheduling algorithm. The algorithm is an extended version of Round Robin Algorithm where each of the processes is given a priority level (low, medium or high) and based on the priority level, the Time Quantum for that process is decided and executed. The throughput of the algorithm is increased by executing processes with smaller burst time or smaller remaining burst time by allocating the process to the CPU as soon as it is ready but not pre-empting the running process.

Abstract: We devise a system for measuring influence of Twitter users, which we call InfluenceRank, based on certain features extracted from their Twitter profiles and tweets authored over the duration of two months. As in the real world, influence of a user on social media may be judged by the engagement they drive through the content they publish. For a tweet, engagement can be most obviously measured by the number of retweets (RTs), favourites and replies it gets. Our system comprises of a regression based machine learning approach with InfluenceRank as the predictor variable against the set of our proposed features. The regression model has shown reasonable accuracy despite being fit on limited data.

Abstract: Criminal activities are a manifestation of unseen termites that are slowly but steadily decaying the deep rooted pillars of ethics and values established in our society. The evolving technology can be very well be utilized as arms and ammunitions by the law agencies against this social evil of criminalization. In our paper, we propose a novel and unified approach to examine and investigate digital crimes as well as physical crimes. Our model works on the principle of integrating various computerized forensic tools to analyze the reported digital crime and adopts data mining techniques for detecting crime and predicting the criminal in the case of physical crimes. In the first phase the user registered in the system can file a valid case by entering the details of the crime occurred. Depending on the type of crime the case will be evaluated. For detecting and investigating intruder attacks launched on a user's system, a set of digital tools is used and the generated report is sent to the intended user. In the event of a physical crime, k-means clustering algorithm is used to generate crime clusters. Based on the crime location the clusters are diagrammatically represented on google maps. We have further incorporated the use of Naive Bayes classification algorithm for predicting the criminals for a particular crime case based on similar crime activities that happened in the past. If no previous record is found then the new crime pattern is added to the existing crime dataset. Our computerized forensic model aids the victim to amicably cooperate with the law agencies and aims to accelerate the process of crime investigation in order to combat rapidly growing criminal activities.

Abstract: Text summarization combines the process of POS tagging, term frequency and topical analysis. All these together are used to produce insightful summary of the document/documents. The concise version of the text document can be made using the concept of frequency of the terms and inverse frequency of documents. Text summarization is useful for bring the short story of all the newspaper articles, email correspondence or to extract key elements for the search engine. To compact the size, the sentences which are not near to the centroid is not to be considered in the output. To do that, the data which does not relate to the centroid topic has to be pruned. The output consists of only important data useful to the user. Large unstructured data can be converted in such form that can be used for report making, compacting of web pages and review of the book. In this, the summary from documents contains significant information, and is less than half of the original size. The output should be such that it fully satisfies the user's query and understands the answer given to it.

Abstract: In Vehicular Ad Hoc Networks(VANET), vehicles communicate among each other and with infrastructure points by broadcasting safety and non-safety messages in the network. Due to wireless communication, security and privacy are very important issues to avoid threat in the network. Group based vehicle to vehicle (V2V) communication scheme is proposed here which prevents vehicle from threat. To achieve security and privacy goals, we propose one time authentication for group and then V2V communication is done using group symmetric key within group. Our scheme satisfies all security and privacy requirements such as authentication, non-repudiation and conditional traceability. In case of malicious activity, this scheme can trace malicious vehicle which generates a false message. Computation and communication cost is improved as compared and analyzed with other previous schemes.

Abstract: Carver Mead at Caltech in the mid-1980s proposed to build devices based on the organizing principles used by nervous system. He coined these systems as ‘Neuromorphic systems’. They are usually composed of analog electronic circuits in the Complementary-metal-oxide-semiconductor (CMOS) technology. These Neuromorphic circuits aim at emulating biological nervous system and its components in silicon hardware. The emulation of neuron behavior at circuit level is one of the most complex tasks in the development of neuromorphic hardware. The analog implementations of neurons require less area and less power consumption as compared to its digital implementation. Thus, these are the serious contenders for future large scale neuromorphic systems. A number of neuron models and their implementations are reported in literature out of which Integrate and fire neurons are found to be most simple, compact, and highly energy efficient circuits. In quest of implementation of more dense and low power circuits in these analog electronic systems, constant scaling of CMOS technology has opened new avenues. In order to get benefit from the scaled down technology node and the Silicon neuron circuits already available, the option of porting an existing neuron circuit implemented in higher process technology to the 180nm process technology is explored in this paper. We have implemented an existing integrate and fire neuron circuit at 180nm CMOS technology node with circuit simulator and studied the resulting characteristics.

Abstract: Network virtualization(NV) provides containers to separate the tasks but share hardware resources and thus leads to reduced cost, increased network service deployment and efficiency of network. This virtual network is separated from physical hardware and can be used for performing operations like creation and deletion. Software-Defined Networking (SDN) separates the control plane from the data-plane of the switches. SDN contains protocols, e.g., OpenFlow (OF) to remotely control the data-plane SDN switches. So by merging SDN and NV we can create virtual Software Defined Networks (vSDNs).

Abstract: Inverted Index is an important data structure in computer science. It is used to create a mapping between a word and the set of documents in which that word appears. Thus, it is used to store documents per word. Currently, the output of inverted indexing is stored haphazardly in a look up table. Hence traversing through the look up table for fetching indexes requires linear search. The time complexity of linear search is O(n) where n is the number of words whose inverted index has been stored. In this paper, a hash based optimization is proposed for storing the output of inverted index which can reduce the searching time complexity to O(1). Since inverted indexes are quite popular in big data applications like search engines, a MapReduce implementation of the proposed technique is also presented which can be easily implemented in a distributed environment.

Abstract: This paper proposes a multispectral magnetic resonance imaging (MRI) image fusion scheme for improved visualization of anatomical and pathological information of meningioma (MG) brain tumors that combines the contourlet transform and fuzzy statistics. The proposed fusion technique mainly targets the tumor and its surrounding hyperintense (edema) region, which leads to improved brain imaging informatics for radiologists. The developed methodology mainly consists of the contourlet transform for multiscale and directional decomposition, fuzzy entropy for fusing approximation coefficients, and region-based fuzzy energy for fusing detailed coefficients of two input images with the same orientations. Two fusion rules are established here in order to fuse corresponding lower- and higher-frequency subbands of images. The proposed methodology is applied to five various combinations (such as T1-weighted and T2-weighted, T1 post-contrast and T2-weighted etc.) generated from four modalities of MRI images (T1-weighted, T1 post-contrast, T2-weighted, and fluid-attenuated inversion recovery (FLAIR)). A total of 150 MRI images (30 images from each of five combinations) are considered from 20 cases of MG brain tumors. A quantitative evaluation of the proposed method is performed in terms of three performance measures. The performance is compared with that of existing medical image fusion techniques tested on the same dataset. Experimental results show the superiority of the proposed methodology in terms of both qualitative and quantitative measures, which also indicates that fused images contain enriched diagnostic information that can aid the detection of tumors and edema. A fusion of post-contrast T1-weighted MRI images with FLAIR and T2-weighted MRI images provided clinically relevant information.

Abstract: Laboratory classes are integral part of an engineering course. Laboratory sessions are primarily designed to develop proficiency in technical skills, provide an opportunity to place theory in context, develop critical thinking skills and promote enquiry based learning. Laboratory experiences will be paramount in developing our students as independent learners, researchers, critical thinkers and generators of knowledge. There are several reforms need to be implemented to improve student laboratory experiences. This paper suggests some of the methods that can be implemented in engineering institutes. This paper also presents the study of usage of online tools and rubric based laboratory assessment methodology adopted in electronics engineering course.

Abstract: Taking care and maintenance of a healthy population is the Strategy of each country. Information and communication technologies in the health care system have led to many changes in order to improve the quality of health care services to patients, rational spending time and reduce costs. In the booming field of IT research, the reach of drug delivery, information on grouping of similar drugs has been lacking. The wealth distribution and drug affordability at a certain demographic has been interlinked and proposed in this paper. Looking at the demographic we analyze and group the drugs based on target action and link this to the wealth and the people to medicine ratio, which can be accomplished via data mining and web mining. The data thus mined will be analysed and made available to public and commercial purpose for their further knowledge and benefit.

Abstract: Computer crime also popularly known as Cybercrime has escalated to such a huge extent that it is now posing a threat to various industries, educational universities and professional organizations as well. The law enforcement agencies, police departments and crime branch units have acknowledged the upsurge in digital crime cases and they have begun to deploy measures to curb this evil phenomenon. In this paper, we inspect about the awareness of digital crime among the general public and illustrate an overview of Cybercrime, with the motive of highlighting the necessity to restrain the impact of cybercrime all over the world. This paper examines the rise in number of cybercrimes in India and takes into consideration the analytical traits of the offenders who commit such crimes. The paper banks on the information obtained from different sects of our country. The experimentation results depict that the top four cyber crimes committed in the past few years such as Internet frauds, data theft, cyber piracy and crime sex were all spread across the internet. The output reveals that cyber crime not only encompasses the internet but it has already expanded across all communities worldwide. The soaring crime rate is a major concern as it is indicative of the huge amount of cyber crime cases enrolled in recent years. The objective of this paper is to provide some guidelines to cybercrime analysts, government organizations, and educational universities.

Abstract: The conversion of an algorithm to code is still at an early stage. Effective conversion of algorithms mentioned in natural English language to code will enable programmers to focus on logic building and free them of syntactical worries, further it will also aid the visually impaired programmers. Although beneficial, implementation of such a converter encounters numerous challenges like limitations imposed due to semantics of the English language, case frames, etc. In this paper we have introduced an interpreter that is capable of converting algorithms in English to C code whose flexibility of interpretation has been enhanced by using synonyms and by the introduction of a personalised training model whose concept has been outlined below. We have defined the conceptual model along with a user scenario which demonstrates the functioning of our model.

Abstract: Security is one of the major concern in today's era, may be it is for your home, locker, your cabin or a restricted area in any institution. There is a necessity to provide safe access to personal rooms, lockers and avoid intruders has compelled to make security systems more digitally compatible. A digital interface can be used to make it more comfortable for access. Sometimes, it becomes tedious to carry physical key to your locker in case of multiple users. The conventional physical key can be replaced by digital password to open your lock. This research is aimed at developing a fruitful security system which provides secured log access to authorized people with the feature of an email notification to the administrator along with the captured image of the user. An important feature of this system is to enter the password using Bluetooth enabled smart-phone. These systems are profoundly welcomed, since it can be monitored from a remote zone. This project has an advantage over physical lock in term of safety, storage and handiness.

Abstract: E-learning applications must be designed so as to cater the different needs of learners as each learner learns differently. An E-learning system is proposed which is hybrid in nature since it would facilitate different modes for content delivery matching with learner's personality based on the Myers-Briggs Type Indicator (MBTI) and the different levels of learning required by the learner based on revised Bloom's Taxonomy. The learner's profile is initialized according to the results obtained by the learner in questionnaire which determines the learner's personality type and dominant personality type. Once a learner chooses a course to learn and selects its learning level, a quiz on the learner's chosen topic is taken. A quiz after the completion of E-learning course would be taken to examine the improvement in learning capabilities of the learner.

Abstract: Data mining is a technology used in different disciplines to search for significant relationships among variables in large data sets. In this paper, we concentrate on the application of data mining in an educational environment. This study can be used to help teachers to classify students' academic success, along with their determination measured by a grit test, and thus modify their teaching for different groups of students. According to this classification, one can arrange remedial classes or extra tests for the required students. Also students can monitor their growth from semester to semester with the help of the application made.

Abstract: In today's competitive world, every institute needs to keep their academic standard as high as possible. It becomes mandatory for all most all the institutes to maintain the quality in technical education as well as to produce the skilled graduates. In order to produce the skill graduates, the Institute always rely on different programs which is responsible for producing the high calibre graduate. As of now, there is no such application available which will automate at least the process of reducing the clerical work required for preparing the course file to evaluate the Course Outcome with Program Outcome .NBA has laid down the guidelines for each program through the means of rubrics to undergo this evaluation process which implies the accreditation grade to be given from NBA committee. This application allows the faculty to enter the details about their courses in terms of mapping with PO, PSO and Bloom's Taxonomy. The application calculates the PO attainment which helps the faculty about the existing gap which further can be improved in the next semester.Hence such type of application will assist the faculty to reduce their workload regarding the individual course.

Abstract: As the need for more complex control and friendly user interface is increasing, digital control implementation is crucial in modern energy conversion systems and power electronic converters. The Load resonant inverter (LRI) can be controlled to generate a quasi square wave of variable pulse width generated by digital signal processor (DSP). It is used to store the required commands for generating the necessary waveforms to control the frequency of the inverter through proper design of switching pulses. In this paper, a digital controller has been implemented which possess advantages of phase shift control while providing rapid frequency tracking for load resonant applications. The digital control algorithm has been optimized in Embedded C Language and implemented on a floating point Digital Signal Processor TMS320F28335. The DSP alogorithm is experimentally evaluated on load resonant inverter used for an induction heating system prorotype.

Abstract: Even though desktop computers have existed since a long time, the method of typing and feeding input has not changed much. Versions after versions of popular text editors have come, and yet no editor yet has addressed the difficulty that smaller user groups such as senior citizens and handicapped people have while entering input. Also, predictive editors cease to exist in desktop computers even today. This paper explores the use of a new software for input on desktops, which relies on a dynamic predictive algorithm using n-grams and suffix trees to significantly reduce the effort of typing. This paper also compares the newly developed Smart Text Editor to traditional popular editors.

Abstract: Many conventional approaches and treatments being insufficient to restrain the repercussion caused due to disorder and disability have initiated exploration on Virtual Reality for therapy. Experts having done research on efficacy of VR to overcome mental disorders specifically Post traumatic stress disorder (PTSD) and phobias demonstrate the scope of technology to overcome physical illness. Use of virtual environment and exoskeletons have seen far better results in conquering the despondency in individual with paraplegia, that is, paralysis of the legs and lower body. The aim of the current paper is to provide the cognizance of dilemma suffered by the patients enduring PTSD, Paraplegia and Phobias also the technical and rationale specifications and application features of the tools used. Analysis and studies confirm the improvement in welfare of patients accomplished by integrating customizable virtual scenarios to represent relevant context for their treatment without exposing them to real-time scenarios.

Abstract: Supply Chain Management based organizations need to focus on development of strategies based on the services provided by them to stakeholders. To improve supplier chain value, optimal services need to be selected which may become agile based on the organizational factors. Agility can be differentiated at different levels of execution in SCM organization. Agility and leanness philosophies based on the range of products and services are provided by the organization. This paper focuses on the agility for services provided by SCM based organization. Since in every system there is fuzziness in its structure and function. The functional fuzziness can be easily reduced and compared to structural fuzziness. In this paper different methods based on fuzzy decision methods are compared for strategy development. Each service provided by organization is mapped with SCM strategies. From this set of services, optimal service is selected. This paper explains a case study of an organization working on SCM strategies. To find optimal service, different decision methods like Saaty's method, minimization of maximal regret method, AHP method, ranking method, and fuzzy average weighted method are compared to measure relativity of fuzziness in multiple objective frameworks by structuring the system hierarchically. This paper also shows advantages and disadvantage of these method as well as mathematical operation for selection of optimal services based on SCM strategies.

Abstract: FANET is co-operative, reconfigurable, autonomous network of UAVs. The emerging technology of FANET has many applications in civilian and military field like search and rescue mission, surveillance, reconnaissance, monitoring etc. which create temporary mobile ad-hoc platform for communication and data collection. For such applications, main requirements are maintaining and sustaining connectivity with nodes and effective utilisation of limited resources like bandwidth. Further, as the UAVs are moving with very high velocity, their topology changes rapidly and hence connectivity among them may lose. To address this problem, usage of novel technique called compressed sensing is proposed. In compressed sensing method, instead of sampling the signal by Nyquist rate, signal is sampled by Sub-Nyquist rate i.e. sampling rate less than Nyquist rate. Thus, few samples are transmitted and signal can be recovered by using convex optimisation technique. In FANET, number of UAVs are few. Hence, number of frequencies occupied will be sparse as compared to available wide spectrum band. In this paper, sparsity in form of number of UAVs is exploited and compressed sensing is implemented on coordination information which is shared between nodes.

Abstract: VoWiFi or deployment of VoIP over wireless medium can facilitate campus-wide provisioning of IP-based voice service. Determining the performance and capacity limits is essential for proper dimensioning of the service in an enterprise. This paper analyzes the real-time performance of an economical RasPBX based Asterisk system for VoWiFi service in terms of voice capacity and quality. The call blocking probability serves as the voice capacity metric and the voice quality is assessed from network packet losses. Prediction of subscriber capacity from these parameters using Erlang-B model to ensure a scalable system deployment has also been outlined in this work.

Abstract: Today’s computer network architecture generates massive amounts of security log data. The analysis of such huge security log data is very difficult. In addition to correlation of security alerts, we require valid and proper predictive log analysis to track the future attacking possibilities. The existing techniques for prediction take input as some preprocessed or mined data. That’s why, the predictive system becomes time consuming and it affects accuracy rate. So in order to enhance the performance of predictive analysis, the proposed system replaces the data processing phase with event correlation technique. It is based on rule-based approach for processing real-time events and produces attack classes as output for multistage attack detection. It reduces processing overhead and leads to low utilization of memory. The hidden Markov model (HMM) makes use of such highly processed data for future attack prediction which makes the system more efficient. In a direct comparison, we concluded that the inclusion of correlative analysis results improves the prediction performance.

Abstract: Multi-gate FET, a self-aligned structure (FinFET) is one of the most promising device to address leakage issues and short channel effects in deeply scaled CMOS technology nodes. In order to improve the analog performance of the system the figure of merits (FoMs) of device technology should relate to those of circuit level FoMs. In this paper, based on Visual-TCAD 3-D numerical simulations, we analyze the effect of multi-fin structure and resulting impact of various RF parameters like transconductance, output resistance, intrinsic gain, parasitic resistance and capacitance, mobility, early voltage, drain conductance and cut-off frequency which affects the analog behavior of multi-fin FinFETs device.