Sriman Bidhan Baray

Sriman Bidhan Baray

PhD Student | Graduate Research Assistant

Publications

You can also find my articles on my Google Scholar profile.

Journal Articles

EOG‑Based Reading Detection in the Wild Using Spectrograms and Nested Classification Approach

Published in IEEE Access, 2023

This paper explores a novel approach to detecting reading activities from electrooculography (EOG) signals collected in real-world environments. By combining statistical features with deep learning models and employing a nested classification approach, the study significantly improves reading activity detection accuracy to 66.56%, outperforming the baseline performance of 32%.

Automated Measurement of Penile Curvature using Deep Learning based Novel Quantification Method.

Published in Frontiers in Pediatrics, 2023

This paper presents a deep learning-based automated method for accurately measuring penile curvature (PC) using 2D images. By leveraging YOLOv5 for localization, UNet for segmentation, and HRNet for landmark detection, the study achieves a mean absolute error of less than 5° on 3D-printed model images and demonstrates high accuracy when applied to real patient images, offering a promising tool for clinical assessment and hypospadiology research.

Conference Papers

An Empirical Approach for Human Locomotion and Transportation Recognition from Radio Data

Published in UbiComp/ISWC 2021 Adjunct: Adjunct Proceedings of the 2021 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2021 ACM International Symposium on Wearable Computers, 2021

This paper presents a machine learning approach for recognizing eight modes of locomotion and transportation using radio data from GPS, WiFi, and GSM cell towers in a user-independent manner. The study explores feature extraction techniques and model optimization, demonstrating that an XGBoost Classifier with selected features achieves an accuracy of 79.11% and an F1 score of 79.39% on the validation dataset.

Complex nurse care activity recognition using statistical features

Published in UbiComp/ISWC 2020 Adjunct: Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020

This paper focuses on human activity recognition in healthcare by analyzing sensor-based accelerometer data to predict 12 nurse care activities in both lab and real-life settings. The study addresses data imbalance and activity variability, employing noise filtering, windowing techniques, and feature extraction, ultimately achieving 65% accuracy and a 40% F1 score using a Random Forest classifier.