Signal Processing
Signal estimation/ Sensing
Wireless Communication
ML/ DL
Vehicular Communication
IoT
Signal Processing and Communication (SPAC) Lab
Stevens Institute of Technology, NJ, USA
Supervisor: Dr. Hongbin Li
Papers:
Conference paper at CISS'25, JHU, Baltimore (19-21 March, 2025)
DOIs:
Overview:
Problem: Radio astronomy faces strong radio frequency interference (RFI) from satellite mega-constellations, often exceeding the ADC's dynamic range, causing saturation and distortion in observations.
Challenges: Saturated ADC output leads to loss of signal amplitude and phase, rendering conventional interference cancellation ineffective and masking weak astronomical sources.
Proposed Framework: A novel method integrates modulo sampling (self-resetting ADC), adaptive reconstruction, and interference cancellation to effectively handle saturated signals.
Reconstruction Approach: Modulo samples are unfolded back to true signal values, preserving amplitude and phase, allowing traditional interference cancellation and imaging to work again.
Applications: The framework enables accurate spectral estimation using adaptive methods (APES, Capon) and interferometric imaging using CLEAN, even under strong interference conditions.
Supervisor: Dr. MD Farhad Hossain
Research Tools: MATLAB.
Paper: Published with Network and Systems Management by Elsevier. (25th April, 2025)
DOI: https://doi.org/10.1007/s10922-025-09922-3
We propose a new network architecture using Smart Meters as distributed access points of communications, ensuring increased reliability and reduced latency for enhancing V2N communications.
Two distinct algorithms, namely, the maximum SNR (MaxSNR) algorithm and the minimum distance (MinDis) algorithm, are proposed for vehicle-SM associations.
We conduct extensive investigations by developing a MATLAB simulation platform for determining the viability and acceptance of our proposed system model and algorithms.
Various performance metrics, including reliability, latency, and throughput are evaluated under various system settings for validating the effectiveness of the proposed architecture for V2N URLLC operations.
The trivial Base Station-based V2N system model is also simulated and compared to determine the superiority of our proposed system model.
Some figures related to this work can be found below.
Supervisor: Shahed Ahmed
Research Tools:Python, ML, DL.
Paper: Google drive
We build a hardware device to make our own database of fingerprint and blood group data for our project.
We try various image processing method on our raw data including Gaussian Blur, Median Blur, Otsu Threshold, Adaptive Threshold Gaussian, Otsu Threshold to Adaptive to Canny to Binary etc.
In our study we extract some fingerprint features by typically involving the identification and extraction of minutiae points, which are the distinctive characteristics of a fingerprint such as ridge endings, bifurcations, and ridge contours.
We have done the tests in our datasets using 80% dataset for the training purposes and 20% datasets for tests. We got 40-45% accuracy when we did the test considering 4 major groups (A,B,AB and O). But we got better results when we did the test using a binary method (considering two groups at a time).
Further we implement a statistical analysis using a one way anova test for this analysis.
Some figures related to this work work can be found below.