M Tech Dissertations

Permanent URI for this collectionhttp://drsr.daiict.ac.in/handle/123456789/3

Browse

Start date: 2010End date: 2019

Search Results

Now showing 1 - 10 of 470
  • Thumbnail Image
    ItemOpen Access
    DCA : (Data capturing application)
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Batada, Arlene; Banerjee, Asim
    Imagine an institute with thirteen departments functioning under it. Now imagine most of these departments functioning with their own methodology, and most of them using their own software system. Can you think of a scenario where the upper management requires data, where a part of it is available from department ‘X’, another from ‘Y’, and a third from ‘Z’, and with this all these departments sending their Excel sheets to the upper management. Now can you imagine the nightmare it is going to be for management in consolidating the huge amount of data it has received from various departments, and then applying filters so as to obtain the exact data that is required? The situation is going to be worse if along with this, some historical data is also to be consolidated into this set, with the departments complaining that the requested data had already been sent months back, which the management has had got misplaced in its ocean of excel sheets. To solve this practical problem, a software system, named DCA- an abbreviation for ‘Data Capturing Application’ is developed. The system rests on exactly three users:, 1. Super Admin 2. Departmental Admin 1 3. Departmental Admin 2 in the descending order of their privileges. The application serves as a platform where forms can be generated dynamically by Super Admin, where each of them will have an association with one of the departments of the institute. Upon generation of a form and allocation of a department to it, the same will then be accessible to the remaining two admins of that department. Each generated form has a deadline, upon elapse of which. no more modifications in the data are permitted by any of the three users. Generally, the responsibility of feeding in the data lies with Departmental Admin 2. The form is then to be submitted by the Departmental Admin 2 once all the data feeding is complete. This action is to be performed prior to the deadline. Upon submission of the form, no more changes are permitted for Departmental Admin 2. However, in case of any changes thereafter, the same can be made by Departmental Admin 1 up to the deadline. Any changes in the data thereafter require Super Admin to extend the deadline of the form in order to enable Departmental Admin 1 to make those. If Departmental Admin 2 is also to be permitted to make changes in this situation, then along with extending the deadline, the action of unsubmission of the form is also required by the Super Admin. Upon agglomeration of data of various forms, a final report can be generated by the Super Admin with join operations and application of queries available as utilities in this functionality of the software.
  • Thumbnail Image
    ItemOpen Access
    Experiment-based approach to understand delay tolerant network
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Kakran, Bhoomiti; Sasidhar, Kalyan P S
    Dissemination of information in challenged environments such as place with nonavailability of cellular services, post-disaster situations, remote place, disruption min network is often a difficult task. This is due to lack of traditional communication channel which equires end-to-end path to deliver the information using store-forward approach. However, Delay Tolerant Network (DTN) can be useful in such challenging environments as it works even when end-to-end path ceases to exist in the network. It follows store-carry-and-forward approach which incorporated with human mobility can enhance the communication opportunities. Hence, in this thesis a smartphone application called Bluetooth DTN app is developed based on the DTN concept. The experiment is conducted to study how a smartphone based DTN setup would function and how can the use of DTN be justified for challenged environments. Similar scenario is created in Opportunistic Network Environment (ONE) simulator, a widely used simulator to study DTN among the research community to verify the results.
  • Thumbnail Image
    ItemOpen Access
    Design of leaf cell layouts for memory compiler
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Nagaich, Esha; Parekh, Rutu; Agrawal, Yash; Srinath, B.K.
    Digital layouts are designed in such a way that it should have minimum area and hence lesser delay.On the contrary analog layouts are made using best matching technique so as to provide same environment to each transistors.In this project standard cells have been designed in 90nm and 28nm technology.The challenges faced in making layouts in both technology node have been discussed.My work involves design of leaf cells layouts in 28nm technology for memory compiler for given 32 bit memory array.To store the data, there is a need of memory array. user requirement.Integrating memory with designed leaf cells is done such way that on abutment connection will take automatically with the pins.For this most important is pin placement and top level routing.The inputs to the memory compiler are designed in this project which includes the following leaf cell blocks such as Pre-charge and mux block,Sense amplifier block,Input driver circuit,Output driver circuit,Control block and decoders.All the leaf cells are DRC,LVS and compatability clean and are interconnected in the top level.The bit cell of the memory is not designed since the layout designing of the leaf cell will not match the DRC's.because all the fabs will have their own DRC's designing the bit cell of memory.Hence the above SRAM architecture is designed with minimum area and leaf cells abutting. Practically the area of above design came to be roughly 1023.359 um2.
  • Thumbnail Image
    ItemOpen Access
    Emerging On-Chip interconnects for futuristic integrated circuit design
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Shah, Urmi; Agrawal, Yash; Parekh, Rutu
    Performance of VLSI technology is strongly influenced by interconnect delay. Scaling of ICs leads to many signal integrity issues. The demand for high speed, low power and compact chip size have led on-chip interconnects an interesting research area. Subsequent miniaturization in VLSI technology has led performance of interconnect dominate over device. Currently, copper (Cu) is facing several limitations with technology scaling down. A dire requirement to replace the Cu interconnects has arrived. Due to these limitations, the overall performance of Cu interconnect has been affected. Future interconnects materials like allotropes of carbon are suggested to be probable replacement of Cu interconnect as they are prone to problems encountered due to technology scaling down. Also they offer superior properties like high thermal conductivity and current carrying capacity compared with copper interconnects.The advanced on-chip interconnects have been analysed at 22nm technology node. It is analysed from the current research work that graphene interconnects possess better performance than copper interconnects. Various performance improvement techniques like conventional CMOS buffer insertion and Schmitt trigger based buffer insertion techniques have been proposed to mitigate the issue of signal integrity for on-chip interconnects. Even optimization techniques like colliding bodies optimization (CBO) technique have been incorporated for the enhancement of tube density of mixed wall carbon nanotube bundle (MWCB) interconnects. Furthermore, modelling technique like finite difference time domain (FDTD) technique has been suggested which gives a close agreement of 2-3 % maximum delay variation with simulated SPICE results. As the technology shrinks, variation due to temperature, fabrication process and environmental fluctuations grows up significantly. This results in variation in output performance. Variability analysis for graphene interconnects has been proposed with various methods like process corner, parametric analysis and Monte Carlo. Timing uncertainty issues like jitter, eye crossing parameter and eye opening factor have been examined for on-chip MLGNR interconnects in the present research work.
  • Thumbnail Image
    ItemOpen Access
    Design of prominent floating point multiplier using single electron transistor operating at room temperature
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Banik, Sanghamitra; Trivedi, Rachesh; Parekh, Rutu; Agrawal, Yash
    This project work has two main objectives. First is to introduce SET based device in digital logic circuit design. SET based devices has tremendous potential for the exploration to improve the current CMOS based device. The other objective is to implement 8bit,16bit and 32bit floating point multiplier. Performance analysis of and comparison of SET based floating point multiplier has been done with 16nm technology.An efficient floating point multiplier based on single electron transistor is proposed in this work. The aim is to work beyond CMOS technology and current trending research area in Nano-electronics.
  • Thumbnail Image
    ItemOpen Access
    Multi-crop classi cation using hyperspectral data: comparison of classical machine learning and convolutional neural network techniques
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Ghodgaonkar, Abhijeet; Ghosh, Ranendu
    Hyperspectral remote sensing is one of the most exciting elds of remote sensing with the enormous size of data due to high spectral resolution. Hyperspectral data has both spatial and spectral components making a datacube which is very large to process due to a high number of features, a large volume of pixels and a low number of labeled pixels. Classi cation of crops classes using hyperspectral data is challenging due to small-sized classes and spectral similarity between various crops, so di erent pattern recognition techniques are developed to classify the data accurately. In the past, Spectral Angle Mapper and GIS techniques were used land cover classi cation. Classi cation using the supervised machine learning algorithms Support Vector Machine, Random Forest, k-Nearest Neighbors, and Multinomial Logistic Regression are used for classi cation. The performance at various training samples percentages (10%, 30%, 60%, and 80%) is analyzed for all classi ers. The classi ers' parameters are ne-tuned to extract the best decision boundary and better confusion matrix. SVM-RBF was the best classi er with an accuracy of 89.1%, 95.9%,94.8% and 99.8% across Indian Pines, Pavia, Salinas, and AVIRIS-NG datasets respectively In the second part of the thesis, Neural Networks and their various architectures have been analyzed. Convolutional Neural Network(CNN) and Recurrent Network are the main focus for Hyperspectral Remote Sensing Classi cation. The convolution operation helps in identifying neighborhood information and perform spatial feature extraction to get important features or patterns. The recurrent network helps train on a limited number of samples by getting better feature representation from the input data. 1D, 2D, and 3D CNN are the architectures on which the hyperspectral datasets are trained on with overall accuracies of 97.7%, 96.32%, and 98.80% respectively. The classical and deep learning techniques are compared and their results on open datasets and AVIRIS-NG dataset are discussed. 3D CNN deep learning model performance is found to be comparable with the SVM-RBF for classi cation purpose.
  • Thumbnail Image
    ItemOpen Access
    Personalized news recommender system using deep reinforcement learning
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Gangopadhyay, Surupendu Prakash; Majumder, Prasenjit
    News recommender system recommends news based on the user's interests and the content of the news articles. News recommender systems are different from other recommender systems as it always has to recommend latest news articles, essentially making it a item cold start problem. Also the system should take into consideration the changing user preferences. The ratings in such a system is collected as implicit feedback i.e. the user will click on the news item if the recommendation looks interesting to the user. hough we have methods like collaborative filtering, Content Based model, knowledge based methods they are ill equipped to handle the item cold start problem and also they do not capture the changing user preferences. Though techniques using deep learning have been used to solve these problems they suffer from over specialization and do not consider the changing user preferences. In this thesis we propose the use of deep reinforcement learning as a means to recommend news articles by taking into consideration the preferences of user and also adapt as per the change in user preferences. Along with this it also maintains the diversity of the news articles that are recommended to the users.
  • Thumbnail Image
    ItemOpen Access
    ECG-PPG device for real time detection of various cardiovascular diseases
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Vora, Yash; Mishra, Biswajit
    The mortality rate in men and women around the world is increasing every year due to Cardiovascular diseases (CVD). Patients having high probability of cardiac abnormalities have to get it identified as early as possible to treat the disease effectively. The physiological signals like ECG and PPG signals can be used for monitoring and detecting the cardiac abnormalities efficiently. In the proposed work, an ECG-PPG device extracts vital physiological signals from the human body and those signals are processed to detect arrhythmic abnormalities and estimate Blood Pressure (BP). The algorithm for processing the ECG signal extracts the feature points like R-peak, QRS complex, P-wave and T-wave and detects arrhythmic abnormalities. The accuracy of this algorithm depends upon how accurately it detects R-peak. The R-peak detection algorithm was tested with the arrhythmic ECG signals available on the MIT-BIH Arrhythmia Database with a extracting the time domain features from the signal with the help of R-peaks information from the ECG signal and estimating the BP. There are 4 different methods by which BP has been estimated and among them the least Root Mean Squared Error (RMSE) for Systolic Blood Pressure (SBP) is 13.18% and 9.98% for Diastolic Blood Pressure.
  • Thumbnail Image
    ItemOpen Access
    Chatbot platform
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Thakrar, Akash; Tiwari, Saurabh
    There are many platforms available in market for creating chatbots and maintaining them. Chatfuel, ManyChat, Google's Dialogflow are some of the leading platforms which provide such facilities. Yet, there is no platform which provides complete end to end solution with automation facilities, analytics and complete control to developers upon their data. This platform aims at creating such a solution where a developer can have full independence of creating chatbot efficiently and effectively and yet have total control over their user's personal data.
  • Thumbnail Image
    ItemOpen Access
    Study of transactional synchronization extensions (TSX) feature in x86 processor architecture
    (Dhirubhai Ambani Institute of Information and Communication Technology, 2019) Jain, Jiket; Agrawal, Yash; Chowdhury, Prohor
    Improving performance has always been at the center of a new CPU design. Simultaneous Multi-Threading is one of the ways to achieve better performance in terms of CPU utilization. In case of Multi-Threading, threads share data and because of which serialization is needed to fulfill synchronization and consistency of shared data between two threads. Such static serialization may be unnecessary and hence Transactional Synchronization Extension (TSX) provides an optimistic approach to do serialization dynamically when needed. A study of TSX is presented. In addition to a brief description of TSX the report presents the flow of verification and possible ways to verify the complex TSX feature. The results itself provides a proof of complexity of the feature in terms of number of CPU cycles simulated to verify it with respect to only Load Store unit. Further, the report suggests possible improvements over different existing architectures and how the feature verification can be improved.