There have been a lot of advancements in the covid detection system worldwide, however, the common problem that these procedures face is the high expense and time-taking process while going through the old manual procedure of test-and-trial of every single person consecutively. So to overcome these crises and present COVID-check at one's convenience, we have tried to introduce an innovative and contemporary solution by implementing artificial intelligence and deep learning algorithms in the digital-health districts. With numerous scientists claiming that audio sounds generated by the respiratory system can be diagnosed and analysed in order to determine the presence of the disease, we have implemented Automatic Speaker Recognition(ASR) and Speech-Audio analysis that could be pragmatic in the early detection and screening of COVID-19.
For this purpose, our model is based upon the performed analysis of the extracted features of coughing, breathing and speech sounds using the Recurrent Neural Network or RNN as the classifier. It mainly remembers and recognizes the previous data samples thus facilitating the prediction of the future data sequence accordingly. The sound waves for the extraction of features have a set of around 45,000 parameters and approximately 1000 data sets. Regarding the network architecture, we have made use of two LSTM (or Long Short-Term Memory) layers and two dense layers. Thus, constructing a competent neural-network model which will convey the probability of the disease.
- LSTM RNN
- Flask
- HTML
- CSS,JS,JQuery
-
Uploading Audio File
-
Final Prediction
