Research on COVID-19 Infection Detection using Voice Analysis

Considering the recent COVID-19 pandemic, user voice can be used to detect early signs of possible infection. The main idea of the proposed project is to help people detect an infection for COVID-19 without having to leave their homes. The main concern with the spread of coronavirus is people misinterpreting seasonal flu for COVID-19 and in panic rushing to hospitals avoiding the logical solution of social distancing. The voice analysis detection system can answer queries related to the disease and perform complex tasks to detect infection in the user’s voice.

Background

During the 1980’s NSF used voice analyses technology on victims of sclerosis and cerebral palsy. During those times voice recognition programs were limited in nature and hence used by very few certain individuals. But as the technology advanced, voice recognition technology was introduced in almost every field. Today most cellphones are pre-equipped with voice recognition software, making it easy for researchers to obtain data for machine learning practices. Popular companies like Google, Amazon, and Apple have their own voice assistants in their products.

It is no new fact that speech is the most common way of communication between individuals. Researchers from all around the world have been constantly upgrading technology to make it easier for humans to interact with products using voice. Using the same technology, we can implement software that can predict users with infection using voice emotions.

The proposed project is to help predict a possible case of infection which could have been impossible to detect otherwise and to help fight the next pandemic.

Methodology

The implemented software will take user voice as input, save the user voice sample for detection - from COVID patients and from healthy people. The system can easily distinguish the voice between infected and non-infected if there are at least 5,000 voice samples available from each group.

Using machine learning, the comparison from these two groups can detect a difference in speech and emotional behavior. The process takes around a few minutes and requires the following steps:

Submit basic information.

Cough three times.

Count to 10.

Say alphabets from A-Z.

Objectives

The proposed project is supposed to make use of user’s voice in order to combat the situation of COVID-19.

Priority Type Objectives
Functional Technical Study on Speech Recognition library in Python.
Functional Personal Enhance the skills of Python.
Nonfunctional Technical Make speech recognition available for every language and accent.

Results/Findings

Following are research papers written by various authors related to the field of voice recognition technology.

Voice Pathology Classification System Using Machine Learning by IJCSMC Journal.

Cognitive Abilities and Predicting Learning Performances using Machine Learning Models by Panduranga Vital Terlapu

Discussion

As the title of the project suggests, the voice analysis system can help people predict early signs of COVID infection. It can also answer queries regarding COVID-19 statistics and QAs.

The proposed project is an AI agent and hence the application can improve over time and usage. It is intended to collect voice sample data from people and improve its accuracy of prediction. The limitation of the software is that it is intended to be installed on a device that can execute it.

The application is not intended to recognize any linguistic words but only to detect voice emotions to predict infection. The application has a simple interface, to make it as much user-friendly as possible. It is aimed to open, close, minimize, record, and show results.

The application is intended to show results once it has collected sample from the user and matched with its database. It must also save the user sample to help improve its features.

The proposed application must be built on Python since its framework supports almost every operating system. Applications make use of the python speech recognition library to detect voice.

Conclusion

Coronavirus affects breathing patterns and other vital parameters. The AI software analyses a user’s voice and results in a score on the likelihood that the individual has coronavirus based on markers observed from known sufferers.

The software is in its early stages. The software should also not be used as a proper medical test for COVID-19.