INTERSPEECH 2021, an International Conference on speech processing and applications, took place in a hybrid format in Brno (Czech Republic) from 30th August to 3rd September 2021. Adria Mallol-Ragolta from the Chair of Embedded Intelligence for Health Care and Wellbeing at the University of Augsburg (EIHW-UAU), Germany, presented the paper “Cough-Based COVID-19 Detection with Contextual Attention Convolutional Neural Networks and Gender Information” in the framework of the Diagnosing COVID-19 using Acoustics (DiCOVA) Challenge. This was a joint work with Helena Cuesta and Emilia Gómez from the Music Technology Group at the Universitat Pompeu Fabra, Spain, and Björn Schuller from EIHW-UAU.
The presented system extracts salient information from spectrogram representations of cough signals using Convolutional Neural Networks, which are then transformed with a contextual attention mechanism before performing the final classification. An interesting contribution of this work is the investigation of the impact of gender for the task at hand. The models that consider information from the patients’ gender obtain better results, suggesting that gender impacts the automatic detection of COVID-19 from coughs.
For further details, the paper is open access through this link