Dr Nikesh Bajaj

Profile

Nikesh Bajaj is a Lecturer in Data Science, Director of Education, and Programme Director at QMSH, Queen Mary University of London. Nikesh worked as a Research Associate at Imperial College London, in National Heart & Lung Institute (NHLI) from 2021 to 2023, and currently associate with Imperial as an Honorary Research Associate. His research work at Imperial is involved Electrocardiography Imaging (ECGI) inverse problem and quantifying the organisation of electrical activities of the heart. During 2019 to 2021, Nikesh worked as a Research Fellow at University of East London, on a Innovate UK funded project - Automation and Transparency across Financial and Legal Services, in collaboration with Intelligent Voice Ltd., and Strenuus Ltd.  The work was focused on Deception Detection in Conversations using Linguistic Markers, which produced a Patent along with a few publications.

Nikesh completed his PhD at Queen Mary University of London, UK and University of Genova, Italy, in a joint program. His PhD work was focused on Predictive Analysis of Auditory Attention from Physiological Signals – PhyAAt. Nikesh also has a 5.5 years of teaching experience in a university in India, which included courses and labs on Signal Processing.

Nikesh is also a mentor and a consultant (alpha testers) at Deeplearning.ai for courses & specializations offered at Coursera, such as NLP, GANs, Tensorflow, MLOps. Nikesh has a few python libraries, such as spkit, phyaat, pylfsr. 

Links:

Homepage: https://nikeshbajaj.in/
PhD Work : PhyAAt Project Page
Python Library SpKit: https://spkit.github.io

Teaching

QHP4701: Introduction to Data Science Programming

QHP4701 Introduction to Data Science Programming is designed to help students start developing the computer science skills that they will need to successfully manage complex Data Science projects. We will be using Python environments and powerful libraries to represent, process and visualise different types of data. In future modules, you will be using and developing further all the skills that you will acquire in QHP4701 Introduction to Data Science Programming.

QHP5701: Exploratory Data Analysis

QHP5701 Exploratory Data Analysis covers a strong background of signals & systems that lays a foundation for Time-Series Analysis. This module starts with basics of statistics, providing the tools to understand the correct choice of measurements. The foundation of signals and systems covers the classical signal and system analysis skills following convolution and Fourier analysis. At the end, module covers the concepts of time-series analysis. Students will learn to choose and apply correct statistical measures of data, analyse basic signal properties, apply Fourier analysis, and analysis the time-series models.

QHM5703: Principles of Machine Learning

QHM5703 Principles of Machine Learning covers the fundamental concepts, methodology and practical tools necessary to understand, build and assess data-driven models that describe real-world systems and predict their behaviour. This module follows the standard machine learning taxonomy to organise problems and techniques into well-defined families (supervised and supervised learning) and subfamilies. This module focuses particularly to the methodology that is used to identify and avoid common pitfalls and reflect on professional aspects of machine learning. Proving abilities to apply the machine learning methodology to build and rigorously evaluate machine learning systems.

Research

Research Interests:

Nikesh’s research is mainly focused on signal processing and machine learning. He has been working on following projects:

PhyAAt: Physiology of Auditory Attention

PhyAAt project explores the physiological data collected during listening tasks to measure the auditory attention, capturing the modulation of physiology that is associated with level of attention. The project webpage includes the details of experiment design, and all the resources (data, code, libraries) are shared as open source at https:\\PhyAAt.github.io

EEG Artifact Removal Algorithm - ATAR
With availability wireless and easy to use devices, employing EEG in diverse experimental settings has increased in research. This allows us to explore and investigate the diverse phenomenon of human brain. However, recording EEG always comes with undesirable artifacts, and a robust algorithm to remove artifacts from EEG without compromising the loss of information is very crucial. With ATAR, we explore the opportunities to solve this problem and make experimental settings more efficient and useful.

Game, Emotion and EEG

Exploring verity of emotion and cognitive state of mind during a game play, using EEG provides additional insights about Human behaviour and effective game design. With our collaborators, we have designed and conducted several experiments and published our findings.

Computational Models, Inverse Problems, ECGI
In this project, we use computational models and inverse problems for clinical problems focusing cardiac electrophysiology. With our collaborators, we analyse the clinical data by employing signal processing, computational models, machine learning and deep learning techniques.

Deception Detection in Conversations
This project includes the computational linguistic analysis for deception detection.

Publications

Full list of publications can be found on Google Scholar

Papers:

1. Granger causality connectivity analysis of persistent atrial fibrillation dynamics reveals posterior wall mechanistic insights, Heart Rhythm O2(2025).
2. Optimising Beat Selection and Averaging for ECGI to Enhance EGM Reconstruction Fidelity. Computing in Cardiology Conference (CinC) (2024).
3. Automatic and tunable algorithm for EEG artifact removal using wavelet decomposition with applications in predictive modeling during auditory tasks. Biomedical Signal Processing and Control55 (2020): 101624.
4. Deception detection in conversations using the proximity of linguistic markers. Knowledge-Based Systems(2023): 110422.
5. Analysis of factors affecting the auditory attention of non-native speakers in e-learning environments. Electronic Journal of e-Learning3 (2021): pp159-169.
6. Fraud detection in telephone conversations for financial services using linguistic features. 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), NeurIPS, AI for Social Good Workshop.
7. Comparative Evaluation of the EEG Performance Metrics and Player Ratings on the Virtual Reality Games. 2021 IEEE Conference on Games (CoG). IEEE, 2021.
8. Indian sign language recognition. 2012 1st international conference on emerging technology trends in electronics, communication & networking. IEEE, 2012.

Patent:

1. System and method for understanding and explaining spoken interactions using speech acoustic and linguistic markers. U.S. Patent Application No. 17/308,222.

Preprint:

1. Phyaat: Physiology of auditory attention to speech dataset. arXiv preprint arXiv:2005.11577(2020).
2. Deep representation of EEG data from Spatio-Spectral Feature Images. arXiv preprint arXiv:2206.09807 (2022).