A smart android application for Covid-19 detection using cough sounds.
The early detection and diagnosis of COVID-19 and the accurate separation of non-COVID-19 cases at the lowest cost, fast, non-contact diagnosis are among the main challenges in the current COVID-19 pandemic. COVID-19 testing becomes crucial and risky for the person.The COVID-19 was first repor
2025-06-28 16:24:59 - Adil Khan
A smart android application for Covid-19 detection using cough sounds.
Project Area of Specialization Artificial IntelligenceProject SummaryThe early detection and diagnosis of COVID-19 and the accurate separation of non-COVID-19 cases at the lowest cost, fast, non-contact diagnosis are among the main challenges in the current COVID-19 pandemic. COVID-19 testing becomes crucial and risky for the person.The COVID-19 was first reported to affect human life in Wuhan City, in the Hubei province of China in December 2019. Since then, the COVID-19 has spread like wildfire throughout the rest of the world, marking its presence in 213 countries and independent territories [1]. The SARS-CoV outbreak originated in the Guandong province of China and later spread to more than 37 countries worldwide, causing over 8000 infections and around 774 deaths [2]. Meanwhile in April 2022 the numbers have reached 509,711,160 for total cases, +235,610 for new cases and 6,243,959 for total deaths worldwide The main issue is the in-person testing method that puts the medical staff and the person itself at serious risk because there are more chances of getting infected in the hospital or where the in- person testing is being conducted because those places are the hotspots where all people with or without COVID-19 visits. Our project provides a solution to this challenge. We propose designing a Smart Android Application that records cough sound which will classify the subject as Covid positive or Covid-negative through machine learning algorithm [4]. This will aware the person about being infected with COVID-19 so as not to interact socially with people, which can prove to be efficient in reducing the spread of COVID-19 virus. We Aim to implement this Application in the medical health Sector to overcome the cost of Testing and to reduce the risks of medical staff by avoiding the in-person testing. We will release the app on Google play store so that everyone should be able to benefit from this Smart App from their comfort zones.
References:
[1] Coronavirus disease (COVID-19) Pandemic, Apr. 2020, [online] Available: https://www.who.int/emergencies/diseases/novel-coronavirus-2019.
[2] Centers for Disease Control Prevention (CDC), Dec. 2017, [online] Available: https://www.cdc.gov/sars/about/fs-sars.html.
[3] Brown, C., Chauhan, J., Grammenos, A., Han, J., Hasthanasombat, A., Spathis, D., ... & Mascolo, C. (2020, August). Exploring automatic diagnosis of covid-19 from crowdsourced respiratory sound data. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (pp. 3474-3484).
[4] Imran, A., Posokhova, I., Qureshi, H. N., Masood, U., Riaz, M. S., Ali, K., ... & Nabeel, M. (2020). AI4COVID-19: AI enabled preliminary diagnosis for COVID-19 from cough samples via an app. Informatics in Medicine Unlocked, 20, 100378.
[5] Coswara COVID cough audio dataset by IISc Bangalore for training and testing. https://coswara.iisc.ac. in [Last accessed on 22-11-2021]
Project ObjectivesThe following are the objectives to be achieved.
- The main objective is to develop a classifier based on machine learning.
- To develop an android-based application which can record cough sounds and classify as covid-positive or covid-negative.
Our Smart Android Application for Covid-19 Detection model contains a Classifier which is trained by machine learning on the basis of the available data set [5] [2] and we will use neural networks [1] and Mel-frequency cepstrum coefficients to extract the features from the coughing audio signal [4]. The Model will be trained on the jupyter notebook and that will be converted to TFlite so that it can be deployed in Android Application. Once our application is ready, the user needs to press the start button, application will open the microphone for 3 seconds and record an audio clip of cough sample. The data will be uploaded to servers. [4] Our backend server converts the audio clip to spectrogram and then apply our machine learning model to classify and give the result whether the sample is of Covid-positive or Covid-negative.
Model Architecture:
Our Project is essentially zero variable cost for testing Covid as compared to other methods such as RTPCR.
It is a time saving approach give results in minutes as compared to other testing methods, which generally takes 1 day or more.
It reduces Decrease in spread due to less interaction as the testing through app will be done at Home.
Technical Details of Final DeliverableAt the end of our project, we expect to have android Application which can record cough sounds and backend server classify those audio cough signals whether they are of covid or non-Covid. The technical details of our project lies in the artificial intelligence and Machine learning through which we trained our Model.
Application Working:
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| PyCharm License | Equipment | 1 | 30000 | 30000 |
| Graphics Card 2Gb | Equipment | 1 | 10000 | 10000 |
| Android Tablet | Equipment | 1 | 30000 | 30000 |
| Transport | Miscellaneous | 4 | 1000 | 4000 |
| Miscellaneous | 6 | 1000 | 6000 |