Publications

The impact of SARS-CoV-2 infection on respiratory health has been shown to depend on the strain of the virus. This paper is the first to probe the impact by analyzing respiratory sound signals.

Debarpan Bhattacharya, Debottam Dutta, Neeraj Kumar Sharma, Srikanth Raj Chetupalli, Pravin Mote, Sriram Ganapathy, Sahiti Nori, Sadhana Gonuguntla, Murali Alagesan

ISCA Interspeech (2022)

We designed a machine learning tool to screen for COVID-19. The tool, named Coswara, records the breathing, cough, and speech sound samples, analyzes them and ouputs a COVID-19 score.

Debarpan Bhattacharya, Debottam Dutta, Neeraj Kumar Sharma, Srikanth Raj Chetupalli, Pravin Mote, Sriram Ganapathy, Chandrakiran C, Sahiti Nori, Suhail K K, Sadhana Gonuguntla, Murali Alagesan

ISCA Interspeech Show & Tell (2022)

Use it

We conducted a first-of-its-kind challenge inviting researchers from around the world to address the binary classification problem of COVID-19 detection using sound samples. Twenty nine teams participated and posted encouraging results on the leaderboard.

Neeraj Kumar Sharma, Ananya Muguli, Prashant Krishnan, Rohit Kumar, Srikanth Raj Chetupalli, Sriram Ganapathy

Elsevier Computer, Speech & Language (2022)

Challenge Portal

While listening to speech, does our ability to understand a spoken language impact our attention to acoustic attributes of the talker? In this paper, we explore answer this question by analyzing data obtained from a behavioral listening test.

Neeraj Kumar Sharma, Venkat Krishnamohan, Sriram Ganapathy, Ahana Gangopadhayay, Lauren Fink

JASA Express Letters (2021)

Co (COVID) + Swara (sound), a unique dataset created to enable exploration on how COVID-19 impacts respiratory sound signals.

Neeraj Sharma, Prashant Krishnan, Rohit Kumar, Shreyas Ramoji, Srikanth Raj Chetupalli, R Nirmala, Prasanta Kumar Ghosh, Sriram Ganapathy

ISCA Interspeech (2020)

How good are humans at detecting change in talker while listening to spoken converstations? How do machine compare with humans in this task? We explored this topic here.

Neeraj Kumar Sharma, Shobhana Ganesh, Sriram Ganapathy, Lori L Holt

JASA (2019)

Inside Science, IISc Pick.

Speech signals contain a fairly rich time-evolving spectral content. Accurate analysis of this time-evolving spectrum is an open challenge in signal processing. Towards this, we designed an approach which overcomes some of the challenges using foundational concepts in signal processing.

Neeraj Kumar Sharma, Thippur V Sreenivas

Elsevier Speech Communication (2019)

Demo

In real-world speech converstation recordings there are lot of instances at which multiple talkers are speaking simultaneously (that is, overlapping speech segments). Applications such as conversation analysis or recognition are adversely impacted by overlapping speech segments. An automatic detection (and annotation) of overlapping segments can be utilized to improve the performance. This paper proposes a solution.

Neeraj Sajjan, Shobhana Ganesh, Neeraj Sharma, Sriram Ganapathy, Neville Ryant

IEEE ICASSP (2018)

Can there be a mathematical model for the beautiful 2-D patterns in a speech spectrogram? Yes! We proposed a multi-component AM-FM model here and evaluated the performance using analysis by synthesis.

Jitendra Kumar Dhiman, Neeraj Sharma, Chandra Sekhar Seelamantula

ISCA Interspeech (2018)

To save time and still grasp the maximum content, we usually speed-up or slow-down a video. In this paper, we discuss the issues associated with audio signals when scaling the speed of video. We propose a new method which preserves the perceptual quality of audio during such scaling operations.

Neeraj Sharma, Shreepad Potadar, Srikanth Raj Chetupalli, TV Sreenivas

IEEE NCC (2017)

Demo

In contrast to uniform Nyquist-rate sampling for processing time-varying signals, we here proposed a non-uniform sampling and analysis method. We demonstrate its application for estimating amplitude and frequency modulations embedded in time-varying sinusoidal signals.

Neeraj Kumar Sharma, Thippur V Sreenivas

Elsevier Signal Processing (2015)

By listening to the horn of a train we easily detect if the train is approaching or receding. Here we design a simple signal processing method to do the same by modeling the Doppler effect embedded in the recorded sound signal.

Neeraj Sharma, Sai Gunaranjan Pelluri, Thippur V Sreenivas

IEEE ICASSP (2015)

Can a trigonometric polynomial, or sum of sine waves, be sampled and reconstructed from its level crossing and extrema samples? Yes! Check this paper to know more!

Neeraj Kumar Sharma, TV Sreenivas

IEEE SPCOM (2014, chosen in top 3)

We explored reconstructing a signal from its level crossing samples. Treating the problem as underdetermined, we make use of ideas from sparse signal recovery.

Neeraj Sharma, Thippur V Sreenivas

IEEE ICASSP (2012)