Research Activity

Signal Processing and Machine learning

In this work, we worked on classification of Media Files based on their location of recording, using Supervised Machine Learning algorithm (support vector machine). We employed a signal extraction algorithm, Fourier Analysis, feature extraction and SVM to classify the origin of recording of audio files, exploiting power signatures embedded in them from the location of recording. Moreover, we managed to introduce a novel feature that increased the classification accuracy by a significant margin.

Designed a designated device for capturing sample audio data file for ENF at our locality.

We ranked 11th among 52 teams from 23 countries. The details of the competition are delineated in the IEEE SP Cup 2016section of this website.

Peer Reviewed Publication From this work in the Conference: 

1. IEEE RHTC 2017 Conference Proceedings 

Title: Power file extraction process from Bangladesh grid and exploring ENF-based classification accuracy using machine learning.

2. IEEE ICEECS 2018 Conference Proceedings

Title: ENF Based Machine Learning Classification for the origin of Media Signals: Novel Features from Fourier Transform Profile

Image classifier implementation

"CNN-based approach to classify cricket bowlers"

Focus: Collaborative research on using existing image classifier to build and Novel Image Classifier model to detect Bowling (Cricket) actions of different players.

1. Built a comprehensive database for different cricket bowlers' action

2. Built a model for image classifier - includes transfer learning for comparison

Publication from this project:

Title: A CNN-based approach to classify cricket bowlers based on their bowling actions 

Undergrad Thesis

"Forward Handover in LTE HetNet"

Implemented a novel forward handover scheme for LTE heterogeneous network, which allows Pico to Macrocell handover process to be comparatively faster. As a result, it improves respective user devices to achieve higher minimum received power at the cell edge.