This talk describes the recent advances made on camera-based respiratory-assessment technology. The need for such methods is established through a couple of industrially relevant problems. Specifically, the problem of estimating the respiration pattern and rate using a consumer-grade 2D camera is considered. The problem is formulated as that of blind-deconvolution of a noisy single-input multiple-output linear time invariant system with periodic excitation. It is then solved using a novel method involving polynomial sub-space projection and selective ensemble aggregation of pixel time-series. The validity of the proposed method is demonstrated through theoretical results and experiments on real-life data. The talk concludes by extending the method beyond perfectly periodic excitation and describing the applicability of the method to a broader class of problems in other areas of signal processing.
Dr. Prathosh obtained his B.E. (July 2011) in electronics and communication engineering from SJCE, Mysore and Ph.D. degree from the Indian Institute of Science Bangalore (June 2015) in signal processing. He was a research scientist in Xerox research center India (XRCI) between September 2014 to November 2016. At XRCI, he was leading the group focusing on non-contact human-body vital measurement, using novel methods in imaging, statistical signal processing and machine learning techniques. He has published research articles in IEEE journals and conferences. He is a co-inventor in 12 U.S. patent applications (one granted). He was a recipient of the KVPY fellowship, TCS research fellowship and MHRD fellowship given by Govt. of India. His research interests broadly span multimedia signal processing, machine learning for temporal data analytics, speech and natural language processing, computer vision and image processing for healthcare analytics and affective computing. He is also a serious student of Sanskrit and philosophical sciences.