Resource-Aware Identification of COVID-19 Cough Sounds Using Wavelet Scattering Embeddings

Abstract

Over the past year, there has been growing inter-est in the potential of artificial intelligence in addressing the ongoing COVID-19 pandemic. Within this context, this paper presents a novel method for a resource-aware identification of COVID-19 cough sounds using wavelet scattering embedding. The proposed method aims to alleviate some of the limitations of traditional deep learning-based classification approaches in resource-constrained settings. Experiments were conducted to demonstrate the ability of the proposed method to differentiate among three types of coughs: Those from COVID-19-related, asthma-related, and healthy cases. Despite the inherent simplicity of the proposed method, compared to related deep learning-based approaches, state-of-the-art performance has been demonstrated. The proposed method was evaluated using both a crowdsourced and clinically controlled dataset. Over all of the experiments, the proposed method achieved an average accuracy of 97.5%, with an average sensitivity of 97.1% and an average specificity of 98.11%.1