Arun Kumar Singh

M.Tech CSE Grad , Indian Institute of Technology Jammu
image

I am a M.Tech Grad student at Indian Institute of Technology Jammu (J&K) with major in Computer Science and Engineering and specialising Information Security. My favourite subjects are Algorithms and Data Structure, Operating Systems and Computer Systems & Networks. My research intrest lies in Machine Learning and Deep Neural Network applications in the field of Information Forensics and Data Security. Currently I am working on the Machine Learning and Deep Neural Net application in Audio Forensics of AI synthesised content. I am open to work on different projects related to software devlopment or any project that lies in the area of my interest. I am always keen to discuss and contribute in new ideas and projects out of my intrest areas also. Feel free to connect with me for projects and research topics.

Work Experiences

Teaching Assistant : Python Programming & Computer Architecture

IIT Jammu | July 2019 - Present

Assisting and Conducting Lab of Python Programming Subject. Assistance in Teaching of subject Computer Architecture. Below are the few responsibilities assigned:

  • Setting up Machine Environment and Clearing doubts during the Lab sessions for Python Programming.
  • Preparing Assignments and Evaluating Lab Practicals for Python Programming.
  • Preparing Course/Study material for Computer Architecture Subject.

Research Intern : Project

IIT Bombay | Dec 2019 - Jan 2020

Worked on Real Time traffic classification over network of IIT Bomabay. Applied machine learning algortihms to classify data packets for the estiamtion of Traffic Analysis over current network without Deep Packet Inspection. Few of the tasks associated with it are:

  • Capturing large amount of Data Packets over the netowrk using Wireshark to train Machine Learning Model.
  • Created labeled data set of all the packets and trained various machine learning model for classification of packets in MATLAB. Choose the best one for real time trafiic classification.
  • Created Python program to capture packet in real time and tested that packet over the chosen trained model.

Teaching Assistant : Web Technology

DA-IICT | Jan - June 2020

Assisting in Teaching and Conducting Labs for the subject of Web Technology (MScIT Course) at DA-IICT, Gandhinagar. Below are the few responsibilities assigned:

  • Setting up envrionment on systems for the Practicals at Labs.
  • Clearing doubts related to subjects in Lab Hours.
  • Evalution of term paper and lab practicals.

Projects

Human and AI Synthesised Speech Dection by Cepstral and Spectral Analysis

Machine Learing & Audio Processing

Classify the Human speech and AI Synthesised speech based on statistics of features extracted. The classification is performed using Machine Learing Techniques. You can see this project here. The features are extracted by perfoming audio processing on the speech signals like calculating Fourier Transform and Bicoherence. Also performed Mel Frequency Analysis. Following are the few major tasks of the project :

  • Collecting the Human Speech and AI Synthesised TTS(Text-to-Speech) Samples and converted them into labelled set.
  • Normalising all the parametrs of Audio to the Same Scale like Mono Channel, Samilar sampling rate and equality in length.
  • Extracted features from Fourier transform, Bicoherence and Mel Cepstral Analysis. Total of 14 features extracted and a structured data set (Traing as well as Testing) along with labels is created.
  • Trained various Machine Learing Models like Quadratic and Linear SVM, Logistic Regressin and KNN, over the trained data set. Choose the Best one for Prediction.
  • Test the model accuracy by 5 Fold Cross Validation on trained data with average accuracy of 96.1 %. Predicted the result on test data and achived the accuracy of 97.56 %.

In this whole project and all experiments and processing is performed on MATLAB Software and Language. Machine Learning models are also trained and tested in MATLAB itself. However for collecting speech samples from different sources few different tools were used.

Finder & Reminder WebGL App

Virtual Simulation

Generating virtual 3D environment for physical world and reminding task associated with real objects or places. The project is completed using Unity 3D and motion is scripted using C#. The whole application is converted into WebGL application and deployed on Microsoft Azure cloud live. You can visit this application here.

Image Text Identification using OCR Tess4J

SIH Project

Scanning of hard copies of good’s train wagon receipt, identifying text via OCR image text recognition library in Java (Tess4J). The Project is done for Ministry of Steel,Government of India under Smart India Hackathon. Following are the major tasks performed:

  • Designed a web application that stores images of wagon receipt on SQL database over Amazon cloud using AWS RDS.
  • Connected AWS SQL aatabase using MySQL Workbench for DBMS managing.
  • Retrieved text from image obtained from database, using OCR in java application and storing back clean only text to MySQL database on AWS Cloud.