讲座题目：Computer Vision based Heart Rate Estimation using Facial Videos
主讲人：Umapada Pal 教授
主持人：吕岳 教授
开始时间：2023-11-21 15:00
讲座地址：闵行校区信息楼133会议室
主办单位：通信与电子工程学院
报告人简介：
Umapada Pal received his PhD from Indian Statistical Institute in 1997 and he did his Post Doctoral research at INRIA, France. From January 1997, he is a faculty member of Computer Vision and Pattern Recognition Unit (CVPRU) of the Indian Statistical Institute, Kolkata. He was the former Head and at present he is a Professor of CVPRU. He is also an Adjunct Professor of University of Technology Sydney, Australia. His fields of research interest are towards different pattern recognition and computer vision problems like Digital document analysis, Camera/video text processing, Biometrics, Image retrieval, Keyword spotting, Video analysis, Medical image analysis, Pose estimation, Image/video generation etc. He has published 525 research papers in various international journals, conference proceedings and edited volumes. Because of his significant impact in the Document Analysis research, in 2003 he received “ICDAR Outstanding Young Researcher Award” from International Association for Pattern Recognition (IAPR). In 2005-2006 Dr. Pal has received JSPS fellowship from Japan government. Dr. Pal has been serving as General/Program/Organizing Chair of many conferences including International Conference on Document Analysis and Recognition (ICDAR), International Conference on Frontiers of Handwritten Recognition (ICFHR), International Workshop on Document Analysis and Systems (DAS), Asian Conference on Pattern recognition (ACPR) etc. International Conference on Pattern recognition (ICPR) will be in first time in India in 2024 under his leadership. Also, he has served as a program committee member of more than 60 international events. He has supervised 17 PhD students. He has many international research collaborations and currently supervising Ph.D. students of 5 foreign universities. Also, he has visited more than 30 countries for his academic work. He is the In-Charge of the joint research cluster of the Indian Statistical Institute and the University of Technology Sydney, Australia. He is the founding Co-Editor-In-Chief of Springer Nature Computer Science journal. He is serving as Associate Editor of many journals like Pattern Recognition, ACM Transactions of Asian Language Information Processing, Pattern Recognition Letters (PRL), International Journal of Document Analysis and Recognition, IJPRAI, and IET Biometrics. Also, he has served as a guest editor of several special issues. He is a Fellow of International Association for Pattern Recognition (IAPR), the Asia-Pacific Artificial Intelligence Association (AAIA), Indian National Academy of Engineering (INAE), West Bengal Academy of Science and Technology (WAST) etc. He is the IAPR fellow selection committee Chair for 2022-2024. Also, he is among the top 2% scientists in the world as listed by the Stanford University in from 2020.
报告内容：
Heart Rate (HR) estimation using computer vision techniques in adverse situations, such as arbitrary face movements, complex backgrounds, occlusion, color changes, lighting effects, etc. is challenging for the video technology community. Unlike state-of-the-art models that use color and spatial-temporal information, the present work exploits vital information in multiple domains, namely, spatial-temporal and frequency-temporal using wavelet transform for heart rate estimation. To achieve this, we explore pyramidal structure to extract invariant features to the above-mentioned challenges from spatial-temporal-frequency domains. This work introduces a transformer to derive context information from multiple domains. Furthermore, to strengthen the features, the proposed model introduces a new attention approach called Mutual-Sharing-Multiple Domains (MSMD) to exchange context information across domains in an end-to-end fashion and it will be discussed in this talk. Experimental results on four standard datasets, namely UBFC-rPPG, VIPL-HR, OBF, and MMSE-HR show that the proposed model is generic and invariant to the aforementioned challenges. Furthermore, comparative studies demonstrate that our method is better in terms of MSE and SD measures. Some demonstrations will also be shown to the participants for better understanding.
