Publications & Patents

Research contributions in eye tracking, computer vision, and machine learning technologies

Research Publications

Academic papers and conference proceedings in computer vision, deep learning, and human-computer interaction

CSA-CNN: A Contrastive Self-Attention Neural Network for Pupil Segmentation in Eye Gaze Tracking

Conference Paper

Algorithm to detect pupil in eye images using a deep neural network with contrastive self-attention mechanisms for improved accuracy in eye gaze tracking applications.

Authors: Soumil Chugh, Juntao Ye, Yuqi Fu and Moshe Eizenman
Conference: Symposium on Eye Tracking Research and Applications

Detection and Correspondence Matching of Corneal Reflections for Eye Tracking Using Deep Learning

Conference Paper

Novel deep learning approach for detecting and matching corneal reflections in eye tracking systems, improving accuracy and robustness of gaze estimation algorithms.

Authors: S. Chugh, B. Brousseau, J. Rose, and M. Eizenman
Conference: 25th International Conference on Pattern Recognition (ICPR)
Year: 2021
Pages: 2210-2217

An Eye Tracking System for a Virtual Reality Headset

Master's Thesis

This research describes a head-mounted eye tracking system for a commercial virtual reality (VR) system. The system uses infrared LEDs that illuminate the eyes and cameras to capture eye images and a 3D gaze estimation model that uses the locations pupil center and corneal reflections in the eye images. It generates gaze estimates that are insensitive to headset slippage. A key contribution of the work is a novel method to determine the correspondence between the corneal reflections and the LEDs using a fully convolutional neural network (CNN) based on the UNET architecture, which correctly identifies and matches 91% of reflections in tests. The eye tracking system has an average gaze accuracy of 1.1°, which is at least 100% better than current VR eye tracking systems.

Author: Soumil Chugh
Institution: University of Toronto (Canada)
Year: 2020

Exudates Segmentation in Retinal Fundus Images for the Detection of Diabetic Retinopathy

Journal Paper

A simpler automated approach to detect diabetic retinopathy taking exudates into account. The method uses homogeneity of healthy areas rather than unhealthy areas, first extracting healthy areas such as blood vessels by entropy thresholding method and optic disc using sobel filter method, then employing thresholding to segment the exudates. The results show that the presented method performs better than previous proposed methods for segmentation of exudates.

Authors: Soumil Chugh, Jaskirat Kaur, Deepti Mittal
Journal: International Journal of Engineering Research & Technology (IJERT)
Year: 2014
ISSN: 2278-0181

Low Cost Calibration Free Pulse Oximeter

Conference Paper

A pulse oximeter device that measures pulse rate and blood oxygen levels in a non-invasive way without any need of calibration. The device uses digital computation power of the microcontroller leading to a simpler circuit design. Two Photo-plethysmographic (PPG) signals corresponding to Red and Infrared wavelengths are obtained from the sensor using an Arduino microcontroller, with signal processing carried out using MATLAB. Oxygen levels are determined using Beer Lambert Law while Pulse rate is calculated both in time and frequency domain.

Authors: Soumil Chugh, Jaskirat Kaur
Conference: 2015 Annual IEEE India Conference (INDICON)
Year: 2015
Pages: 1-5

Non-invasive Hemoglobin Monitoring Device

Conference Paper

A non-invasive hemoglobin-monitoring device that measures hemoglobin concentration without utilizing a drop of blood. The method uses the principle of photoplethysmography and Beer Lambert law to measure hemoglobin levels. Two PPG signals are obtained by shining lights corresponding to Red and IR wavelengths on the fingertip, and absorbance levels are computed. Path length is determined using the refractive index of hemoglobin and physical distance between source and detector, removing the need for any calibration.

Authors: Soumil Chugh, Jaskirat Kaur
Conference: 2015 International Conference on Control Communication & Computing India (ICCC)
Year: 2015
Pages: 380-383

Effect of Different Signal Processing Techniques on a Calibration-Free Pulse Oximeter

Conference Paper

Investigation of various signal processing techniques to improve the accuracy and reliability of calibration-free pulse oximetry systems for medical applications.

Authors: S. Chugh and A. Akula
Conference: 3rd International Conference for Convergence in Technology (I2CT)
Year: 2018
Pages: 1-6

Patents

Intellectual property in eye tracking, computer vision, human-computer interaction, and healthcare technologies

Multi-modal interaction for selecting semantic regions in agent-based image editing

Patent

Interaction methods for agent-based image editing using multi-modal input for enhanced user experience and precise semantic region selection.

Inventors: Yu Zhao, Abby Lu, Soumil Chugh, Che Yan, Yuan Deng
Status: Submitted

Methods for cross-media configuration on virtual keyboard theme

Patent

Advanced interaction methods for cross-media configuration and virtual keyboard theming in modern computing environments.

Inventors: Yu Zhao, Yuan Deng, Soumil Chugh, Che Yan, Wei Li
Status: Submitted

A method for cursor positioning with gaze-assisted pen interaction

Patent

Innovative method combining gaze tracking with pen interaction for precise cursor positioning and enhanced user interface control.

Inventors: Juntao Ye, Shuyang Teng, Soumil Chugh, Andy Chow, Yu Zhao
Status: Submitted

Methods and systems for multi-camera corneal reflection gaze system

Patent

Advanced algorithm utilizing multiple cameras and single corneal reflection for precise gaze estimation in eye tracking systems.

Inventors: Soumil Chugh, Juntao Ye and Moshe Eizenman
Status: Submitted

Methods and Systems for Gaze Tracking using One Corneal Reflection

Patent

Innovative algorithm that uses reflection from only one light source to estimate gaze direction, simplifying eye tracking hardware requirements.

Inventors: Soumil Chugh, Juntao Ye and Moshe Eizenman
Patent #: 92023266PCT01

Methods and Systems for Gaze Tracking and Gaze Tracking calibration

Patent

Advanced algorithm using a 3D model approach with no corneal reflection for robust gaze estimation and calibration in eye tracking systems.

Inventors: Soumil Chugh, Juntao Ye and Moshe Eizenman
Patent #: 92023211US01

Prediction of blood renal diseases using digital biomarkers

Patent

Innovative algorithm that predicts renal diseases using digital biomarkers collected from smartwatch devices for early detection and monitoring.

Inventors: Mike, Soumil, Javi, Sean
Patent #: WO2025106664A1

Methods and Systems for Gaze Estimation

Patent

RGB-based eye tracking system that uses the significant eye of a person to determine gaze direction without requiring specialized hardware.

Inventors: Soumil Chugh, Juntao Ye
Patent #: 92005506US01

Methods and systems for gaze-assisted interaction

Patent

Eye tracking system designed for assisted device applications, enabling enhanced accessibility and interaction for users with disabilities.

Inventors: Soumil Chugh, Juntao Ye and Manpreet Singh
Patent #: 92026995US01

Quality control of user‑generated biological sample cards

Patent

Computer vision algorithm to analyze biological sample cards for quality assessment, reducing errors in medical testing and improving diagnostic accuracy.

Inventors: Michal Depa, Soumil Chugh, Javi, Sean, Sidhant, Theressa
Patent #: WO2023091455A1