Riya Bubna

Harmonising Wi-Fi Direct Scanning Phase with Wi-Fi Scanning Phase

Summer Internship
Samsung Research and Development, Bangalore

Wi-Fi Direct connection is established when two devices go to a separate Wi-Fi direct mode and then probe requests and responses are exchanged. This process is generally slow and consumes additional power. The aim of this project was to harmonise the scanning phase of Wi-Fi Direct with that of Wi-Fi. Hence, the devices in question only needed to be connected to the same access point to establish a peer to peer connection between themselves.

• Summers 16'

Route Choice Modelling

Center for High Performance Embedded Systems (CHIPES)
Nanyang Technological University, Singapore

The aim of this research project was to predict the path chosen by a particular driver to travel from a given point A to a given point B. The choice was affected by different parameters like the traffic encountered, the number of traffic signals, the distance, the time taken and so on.

Report: Route Choice Modelling

• Summer 15'

Autonomous garbage collector and separator

Inter-Hall Hardware Exhibition, IIT Kharagpur

The challenge was to research on the different kinds of garbage that could be distinguished and then design the robot in such a way that separate collection could be achieved. After a lot of study, the final categories were decided to be metallic waste, plastic and otherwise. Metallic waste was detected using a metal sensor. For non metallic waste, a camera captured the image of the garbage and a model was trained to predict if or not it was a plastic waste.

This project is being continued to make the model more accurate and allow for segregration of garbage into more categories.

Report: CleanTrak

• August 15' to March 16'

Comment Analysis

Computer Science and Engineering, IIT Kharagpur

The aim of this project was to study various newspaper articles online and find out those points where the comments were not related to the newspaper article. This involved studying the different topics covered by each article and comparing them with the topics of each comment. Herd mentality was observed - The number of new users engaging in any topic was highest for the topics which already had many comments related to it.

Report: Comment Analysis

• January 16' to April 16'

American Sign Language Recognizer

Computer Science and Engineering, IIT Kharagpur

American Sign Language is one of the most important tools for communication for people with hearing and speaking disabilities. However, these disabilities make it really difficult for them to perform communication tasks with people who are unable to understand their language. To this end, in this term project, I have tried to create a recognizer for this using DataGloves for reading data and have used Machine Learning to train our classification models. There is a comparison for all the common models that have been tried and used for this purpose and also, the accuracy at detection of words after training our models on character and digits as a comparison across all the implemented models.

Report: American Sign Language Recognizer

• August 16' to December 16'
• Github

Scientific Paper Recommendation System

Bachelors Thesis Project, IIT Kharagpur

This project proposes a deep semantic similarity model (DSSM), a special type of deep neural networks designed for text analysis, for recommending target research papers to be of interest to a user based on his research interests. An author’s research interests are found using the papers he has written and the papers that he has cited in his publications. The dataset consists of over a million papers, each entry provides the information of the authors, title, abstract and the citations of the paper. A neural network (more specifically, a DSSM) is trained with this dataset and hence, for any unseen paper as input, it decides if the paper should be recommended or not. The implementataion is in progress.

Report: Scientific Paper Recommendation System

• January 16' to Present

Identification of code switching points

Computer Science and Engineering, IIT Kharagpur

Identified Code Switching points in tweets. Firstly, the points where a switch in language occurs were identified. If a language change was found, it was checked if or not it was a code switched point using Deterministic and Probabilistic Models and the results were obtained with an accuracy of around 80%

Report: Code switching points

• August 15' - December 15'

Mailing System

Computer Science and Engineering, IIT KHaragpur

Developed an end‐to‐end email application which enables the users to send emails to other users and retrieve the emails destined for them. The email application was simpler compared to the existing email applications which provided only the basic functionalities like sending and receiving mails and reading emails. SMTP was used for sending mails and POP for retrieving emails. Also, both the server side and the client side was created.

• March 16'

Online Interactive Teaching Website

Computer Science and Engineering, IIT Kharagpur

Developed an online teaching website which had provisions for students to take different courses, teachers to post notes, assignments and quizzes and even assess them. The database used was MongoDB (with Mongoose ODM). The front-end of the website used HTML, CSS, JavaScript, AngularJS. ExpressJS was used to create the API layer over the database.

Report: Online Teaching Website

• March 16'
• Github

Airplane Management System

Computer Science and Engineering, IIT Kharagpur

An online portal for the management of airlines was created. It had almost all functionalities as present in any airline management sites presently, like booking flights, cancelling, checking all airlines available for a particular date, different classes of seats, different kinds of airlines etc. Also, a secondary user (i.e. a manager) could add/remove arilines, add/remove airplanes for any airline and edit the flight schedule. The front-end was developed using HTML and CSS. the backend used PHP and MySQL database was used to store all data.

Report: Airplane Management System

• April 14'
• Github

Steganography

Computer Science and Engineering, IIT Kharagpur

Steganography ​is the practice of concealing a file, message, image, or video within another file, message, image, or video. In this project, any data was compressed using the zlib compression/decompression library. The compression ratio for huge data is very high which helps store a lot of information in an image without distorting the image. This compressed data was then encoded in the image following a particular pattern. An algorithm to retrieve the data hidden in the images was also created and then this data was decompressed using zlib libraries to obtain the original string encoded.

Report: Steganography

• September 16' - November 16'
• Github

Erudito - Citation Analyser

Inter-Hall Open Software

Erudito is a citation analyser which allows users to efficiently manage papers and also gave features to make notes and bookmark authors and/or journals..

As part of this software, I implemented teh back-end parser which would search websites like Google Scholar and Academia to get the details of research papers, authors, journals and citations.

• March 13'

Registration Portal - Robotix

Technology Robotix Society

Robotix 2013 used a Django application based registration portal as the backend software for event management inlcuding team registration, scoring and round management.

• January 13'

Bus Tracker

Computer Science and Engineering,IIT Kharagpur

Designed an information system to track the buses within IIT Kharagpur campus. On board GPS systems output were provided. The goal was to build an andriod application to get updates about nearest bus, expected arrival time etc on a map to the students. The project was part of a larger project to develop a monitoring system for school kids in metro cities of India with sms/smartphone based alerting mechanism for parents.

• March 16'

Text to Braille Converter

Inter-Hall Hardware Exhibition

Text to Braille Converter was a device which took A4 sheet printed pages over which a CMOS sensor would scroll over to get the image. The image was then converted to a character whose braille was then printed on a platform driven by 30 servo motors. All processing was performed on 8-bit ATMEGA microcontrollers.

I designed the OCR module which used naive but fast method so that it can work on 8-bit microcontrollers, testing it initially using openCV and then shifting to AVR. Data transmission of images was also optimised by converting to Bayer for faster communication.

• March 12'

Hobby Robotics Projects

Tech-Fests and Workshops across India

Since my first year in college, I have built multiple robots as a part of workshops that I have participated in, taught in or just to learn more about robotics.

Some of the robots that I have built are:

• Grid Follower - Robot that can follow a black and white grid using traversal algorithms like Depth First Search.
• Lane Follower - Robot that can follow a lane with sharp turns.
• Person Follower - Using PIR sensors, the robot could follow a human.
• Vibration detector - Robot that avoids vibrations in the ground.

GoSafe

GoSafe is an android application designed keeping in mind the safety of women.

Suggests safe route to travel and halt: Every time a user wants to go one from place to another, GoSafe shows the user two paths - the route suggested by Google Maps API which is the fastest route to reach the destination, and the safest route computed by our safety algorithm. The safety route is computed by calculating the safety index of all possible routes to reach the destination from a certain point and then returning the route with the highest safety index.

Helps user to decide safe place to stay or visit: We parse data from local newspapers using Natural Language Processing to obtain information about Theft, Harassment, Accidents, not well Illuminated Roads, CCTV Cameras, presence of Police. All six parameters are available in the form of markers in the app.

Allows users to share how safe they feel in area.

• January 13'
• Github