Evaluating the COVID-19 Identification ResNet (CIdeR) on the INTERSPEECH COVID-19 from Audio Challenges

TL;DR

This study evaluates the COVID-19 Identification ResNet (CIdeR) model on INTERSPEECH 2021 challenges, demonstrating its effectiveness in diagnosing COVID-19 from cough and speech audio with significant improvements over baselines.

Contribution

The paper applies and assesses the CIdeR model on new COVID-19 audio datasets, showing its potential for accurate diagnosis from speech and cough recordings.

Findings

CIdeR outperforms baseline models in COVID-19 detection.

Effective binary COVID-19 diagnosis from cough and speech audio.

Demonstrates the model's applicability to real-world audio challenges.

Abstract

We report on cross-running the recent COVID-19 Identification ResNet (CIdeR) on the two Interspeech 2021 COVID-19 diagnosis from cough and speech audio challenges: ComParE and DiCOVA. CIdeR is an end-to-end deep learning neural network originally designed to classify whether an individual is COVID-positive or COVID-negative based on coughing and breathing audio recordings from a published crowdsourced dataset. In the current study, we demonstrate the potential of CIdeR at binary COVID-19 diagnosis from both the COVID-19 Cough and Speech Sub-Challenges of INTERSPEECH 2021, ComParE and DiCOVA. CIdeR achieves significant improvements over several baselines.