1

I was doing a machine learning project, assigned by my advisor. The project description is as follows - given an audio sample of a patient, we want to predict whether that patient to whom the audio sample belongs to, is suffering from influenza-like illness (ILI). Unfortunately, I have hit a roadblock at the very start, because I am not able to find any dataset or previous work related to my project. Basically, I want labelled audio samples of people suffering from ILI, if someone has them. I would also love if any of you would redirect me to other forums or sites where I can ask for datasets, and also if you know of any relevant related work. The idea was that people suffering from ILI would have sore throats, thereby having some differences in voice from normal speech, and we would like to see if we can exploit that.

  • I guess this is related: businessinsider.com/… – philshem Sep 22 at 7:56
  • Yes, I think the idea of my project is the same, except I want data for ILI. It has been a while since I have been searching, and I haven't got any leads. – Kushan Shekhar Sep 25 at 4:33
1

As @philshem explained, this is an ongoing research area especially with the COVID situation nowadays and it is a bit tricky at this early stage to find open-sourced datasets. Luckily, there are a couple such:

If those are not sufficient, I would advise you to search for similar projects or papers related to the topic and contact the authors with the hope that maybe, they share their data with you under certain conditions/ licensing. The following projects/ papers should help in that regard:

| improve this answer | |
  • 1
    Really thanks for the help. I haven't checked all of the links in detail as of yet, but most of the datasets consists of coughing and wheezing sounds. What I wanted was labelled datasets that would contain people talking or saying something. The dataset has been really difficult to find, but at least this seems like a ray of hope, I will check them out in detail to see if there's anything that would meet the requirements. – Kushan Shekhar Sep 30 at 6:26
0

This is a hot research topic. Best chance is to contact researchers, although medical audio recordings are going to be highly sensitive, so getting raw data is not so likely.

I found one app that asks users to submit voice samples.

https://www.covid-19-sounds.org/en/

It's funded by the EU, so there may be some initiative to make the data open: https://erc.europa.eu/news/ai-app-part-funded-eu-collects-sounds-covid-19

| improve this answer | |
  • Yes, I was told this was new, but so am I. Seems like the main research topic is doing the same as my project, except it is for covid, I think my guide gave me ILI for starters, however I can't proceed further without data, and I have been looking for a while now. I think I should include covid related audio samples and work on it directly, since they have a higher chance of being available it seems. – Kushan Shekhar Sep 25 at 4:38
  • I wouldn't wait for a dataset. Part of research is usually collecting your own data. – philshem Sep 25 at 8:22
  • Yeah, that was my first try. I tried for around a month, contacted friends, tried to get help from the local hospital but nothing worked. So that's why I am now trying to find pre-collected data for the project because collection of own data looks bleak as well. – Kushan Shekhar Sep 26 at 5:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.