The central research question is whether cough audio can be used to distinguish healthy individuals from those with respiratory illness in general (not limited to one disease) using a lightweight model.

This dissertation targets the following objectives: •General Illness Detection: Develop a cough classiőcation model that detects signs of respiratory illness (vs. healthy), rather than focusing only on COVID-19. In our context, ’illness’ includes any respiratory conditions, including COVID, as inferred from the data labels. •Lightweight Model for Mobile Deployment: Design and train a deep learning model that is small and efficient enough to run in real-time on a typical smartphone, without sacrificing much accuracy. This involves choosing an appropriate architecture and hyperparameters to minimise computational load. •Mixed-Dataset Training for Robustness: Use two complementary datasets (Coswara and CoughVid) to expose the model to a wide variety of cough sounds and recording scenarios. We will document how subsets from each dataset were selected and merged, ensuring balanced and meaningful training data. •Real-World Applicability: Consider practical aspects such as noise in recordings, ease of integration into an app, and how the model could be used as a screening tool. We aim to bridge the gap between an academic exercise and a deployable solution.