Voice-Based Common Cold Detection

The identification of health-related issues without invasive testing constitutes a pivotal facet of medical research and technological advancement. This paper attempts to establish a very efficient and precise means of detecting cold symptoms via voice recordings, by comparing various machine learning methodologies and deep learning methodologies such as Convolutional Neural Networks (CNNs) by analysing alterations in one's vocal patterns. The method involves collecting voice samples from people with and without cold symptoms. During the pre-processing stage, the vibrations and various characteristics in the voice are measured. These characteristics capture the unique traits associated with changes in voice quality caused by cold-related effects which serve as input to the models. Time-tested metrics such as Accuracy, the Confusion Matrix, and the F1-score were used for evaluations. Beyond the quantitative, this study embodies our aspiration to tap into the unique intricacies of voice for proactive cold detection. This approach can lay the foundation for unobtrusive and economically viable health-monitoring tools. To conclude, this paper helps us to detect cold symptoms by a voice recording of a human using machine learning and neural network techniques.