Omer Ihtizaz

Academic

Career

Verbatim text

Covitector 

Covitector is a first ever non-invasive, rapid, free of cost platform to detect Covid through the input 
of a forced cough sound. Cough Sound is the only single feature that every Covid Positive or Covid 
Negative patient can experience, where other symptoms like loss of taste or loss of smell is not 
experienced by all patients. Covitector uses a mix of Deep Learning Techniques to produce some of 
the very complex models to understand the inter-connectivity and correlation of features from 
MFCCs and Delta MFCCs, to form a complex algorithm to diagnose people. There were 6 models 
trained and tested on different data samples that are openly available, conjuring a total of 98.4% on 
CoughVid and 100% on Coswara and Virufy-Covid Datasets.  
A ReactJs application backed by Flask, Google Firebase is built to aid people with a web application, 
where they can record, upload, check daily statistics of Covid Cases across the globe; google-maps-
api integrated with the web app helps customers get live update of Covid-positive cases in their 
vicinity. 
Features include: 
- Recording their Audio Sounds - Uploading pre-recorded Audio Sounds. -Applying Silence Removal 
- Applying Cough Detection Algorithm  
- Computing MFCCs  and Delta MFCCs 
- Majority Voting on the cough sound and notifying the results.  
- Data Scraping from worldometers website for Daily Statistics 
- Integration of google-maps-api to produce live update of covid positive cases in vicinity 
 
Technology Used: 
DLTS, Python, ReactJS,  
Colab, Flask, Firebase, Google Maps Api 
Supervisor Name: 
Mr. Umair Arshad 
Co-Supervisor Name: 
Mr Salman Ijaz (PhD Researcher at NTNU, Norway) 
Group Members:   
Omer Ihtizaz (I18-0404) 
Faaira Ahmed (I18-0423)

AI enrichment