Talk details
Title: The COVID-19 Connection: Exploring breathing, cough, and speech sounds for respiratory healthcare
Speaker: Neeraj Sharma, Assistant Professor, Mehta Family School of Data Science and Artificial Intelligence, IIT Guwahati, PostDoc (CMU), PhD (IISc)
Date: 28 January 2023
Time: 7 PM (IST)
Abstract:
Multiple studies reporting COVID-19 symptoms suggest impairments to the respiratory system. As respiratory sounds, namely, breathing, cough and speech, originate from the respiratory system, it becomes interesting to understand if there are detectable signatures of COVID-19 in the acoustic attributes associated with these sound signals. This understanding can encourage design of audio-based, easily deployable point-of-care screening tests for COVID-19. Such tests can supplement the gold-standard RT-PCR and other molecular testing methods. Exploring this direction, we launched a study encompassing creation of a respiratory sound dataset, acoustic analysis of the data using signal processing and machine learning approaches, and development of a COVID-19 screening tool (https://coswara.iisc.ac.in/). We find that significantly above chance COVID-19 detection performance (AUC-ROC>80%) is obtained using sound recordings, and the performance improves on using multiple sound categories, and augmenting symptoms information. In this talk, I will present this study. This work is a joint work with wonderful people who joined hands, remotely, during the pandemic.
Speaker’s bio
Neeraj Sharma was born in Bhubaneswar, a coastal city in eastern India. After completing his undergraduate in Instrumentation and Electronics Engineering from College of Engineering and Technology, Bhubaneswar, he pursued a Masters and PhD from the Indian Institute of Science, Bangalore, India. His PhD thesis proposed time-varying signal models for sampling and reconstruction of non-stationary signals, and applied them to speech signal analysis, modification, and reconstruction application. In 2017, he was awarded the BrainHub postdoctoral fellowship to explore speech signals from a neuroscience perspective at Carnegie Mellon University (USA). His work on talker change detection was awarded Technical Area Pick for Speech Communication (2019) by the Journal of Acoustical Society of America. In 2021, he was awarded the CV Raman Postdoctoral Fellowship by the Indian Institute of Science, Bangalore. The project titled Coswara, an outcome from work at IISc, is currently available as an open-access website tool for public use (https://coswara.iisc.ac.in/). In 2022, he worked at the International Audio Laboratories, Fraunhofer IIS, Erlangen (Germany) exploring temporal perception of spatial audio signals and use of EEG signals for audio processing. Currently he is an Assistant Professor at the Indian Institute of Technology (IIT) Guwahati (India). His research interest resides at the confluence of signal processing, artificial intelligence, and sensory signals capture and understanding. He is founding the SPIN Lab at IIT Guwahati, which aims to pursue research and development on building intelligence by modeling sensing and perception of various kinds of signals around us (https://spinlabiitg.github.io).