Dibakar Roy

Electrical & Computer Engineering

Dibakar Roy

ECE graduate with research interests in brain-computer interfaces, neural engineering, intracortical speech decoding, EEG signal processing, and biomedical machine learning.

Rajshahi University of Engineering & Technology (RUET), Bangladesh

About

I completed my BSc in Electrical & Computer Engineering at RUET in 2026. My undergraduate research examined parameter-efficient cross-session calibration for intracortical speech decoding and subject-independent EEG workload classification.

I am interested in neural signal processing, brain-computer interfaces, and adaptive neural-decoding methods that remain reliable across people, recording sessions, and sensor configurations. I also enjoy building software and embedded systems that translate technical ideas into usable tools.

During my undergraduate studies, I served as a class representative, supported university and American Center events, and represented my department in competitive sports.

Research interestsNeural decodingEEGBrain-computer interfacesCross-session neural adaptationAdaptive neural decodingBiomedical machine learning
Google Scholar Download CV Email
01

Research

Selected undergraduate research in neural engineering and EEG-based machine learning.

Undergraduate thesis · Intracortical BCI

Rapid parameter-efficient calibration for cross-session speech decoding

A cross-session intracortical speech decoding framework evaluated on the public T12 and T15 datasets. Its input-space calibration strategy updates only 65,792 parameters—less than 0.05% of the complete decoder—while keeping the BiGRU encoder and CTC classifier frozen.

  • Strict chronological training, validation, and held-out future-session evaluation
  • T12 macro phoneme error rate reduced by 11.37% and 15.51% with 40 and 80 calibration trials
  • Up to 10.05% relative improvement on independent T15 evaluation

EEG · Workload classification

Baseline-aware EEG classification

A subject-independent EEG workload classification study using leakage-safe evaluation, reduced electrode configurations, and nested leave-one-subject-out validation.

  • EEGMAT and STEW datasets with nested leave-one-subject-out validation
  • Reduced and shared montages with handcrafted and Riemannian features
  • Balanced accuracy of 0.8199 on EEGMAT and 0.8519 on STEW
02

Publications

Peer-reviewed and accepted work listed in reverse chronological order.

  1. Neurocomputing · Elsevier · 2026

    Rapid Parameter-Efficient Calibration for Cross-Session Intracortical Speech Decoding: Cross-Dataset Evaluation

    Submitted · With Editor

  2. QPAIN · 2026

    Fast Label-Free Cross-Session Calibration for Intra-Cortical Speech Decoding

    PublishedDOI

  3. ECCT · Taylor & Francis · 2026

    Baseline-Aware Practical Framework for EEG Workload Classification with Reduced Electrodes Across Multiple Datasets

    Accepted · PresentedScholar

  4. IEEE ICCIT · 2025

    Hardware-Aware RIS-Assisted ISAC with Quantized Feedback: A Robust DRL Framework

    PublishedDOI

  5. IEEE EICT · 2025

    Learning-Augmented RIS-Aided Hybrid Beamforming for Secure and Low-Latency Transmission

    PublishedDOI
03

Academic and software projects

Coursework and independent projects across web development, embedded systems, and applied engineering.

trECEnto event management platform overview

ECE department event management platform

trECEnto

A responsive event platform with account verification, event registration, announcements, role-based administration, and participation analytics.

Next.jsRBACAnalytics
Nutrition Intelligence website project overview

Software Project I

Nutrition Intelligence

A web application for BMI, BMR, and daily calorie estimation with activity-aware nutrition guidance and OTP-secured accounts.

JavaScriptWebOTP authentication
Women Safety Device project overview

Embedded systems project

Women Safety Device

An ESP32-based device paired with an Android SOS application for GPS sharing, Bluetooth control, and incident logging.

ESP32AndroidIoT
04

Background

Education · 2022–2026

Rajshahi University of Engineering & Technology

BSc in Electrical & Computer Engineering · CGPA 3.39 / 4.00

Technical background

Programming
Python, MATLAB, C++, JavaScript
Neural signals
MNE-Python, EEG preprocessing, spectral analysis
Research tools
LaTeX, Jupyter, Git, reproducible workflows
Development
Next.js, MERN stack, HTML, CSS

Industrial training · 2025

BJIT Academy

SDLC, Agile/Scrum, Python OOP, and software quality assurance.