Attention-based hybrid CNN-LSTM and spectral data augmentation for COVID-19 diagnosis from cough sound

Abstract

COVID-19 pandemic has fueled the interest in artificial intelligence tools for quick diagnosis to limit virus spreading. Over 60% of people who are infected complain of a dry cough. Cough and other respiratory sounds were used to build diagnosis models in much recent research. We propose in this work, an augmentation pipeline which is applied on the pre-filtered data and uses i) pitch-shifting technique to augment the raw signal and, ii) spectral data augmentation technique SpecAugment to augment the computed mel-spectrograms. A deep learning based architecture that hybridizes convolution neural networks and long-short term memory with an attention mechanism is proposed for building the classification model. The feasibility of the proposed is demonstrated through a set of testing scenarios using the large-scale COUGHVID cough dataset and through a comparison with three baselines models. We have shown that our classification model achieved 91.13% of testing accuracy, 90.93% of sensitivity and an area under the curve of receiver operating characteristic of 91.13%.

Explainable COVID-19 detection using fractal dimension and vision transformer with Grad-CAM on cough sounds

“…As seen in Table 4 , Islam et al [16] and Manshouri [18] used the VIRUFY dataset and obtained 93.8% and 95.86% accuracy scores with their proposed methods. Hamdi et al [17] used the COUGHVID dataset and obtained a 91.13% accuracy score. Rahman et al [21] used CAMBRIDGE and QATARI datasets and obtained a 96.5% accuracy score.…”

“…The experiments on Virufy dataset yielded a 93.8% accuracy score for mixed features from both the time and frequency domain. Hamdi et al [17] used an attention mechanism-based CNN-LSTM approach for cough-based COVID-19 detection. Authors used a spectral-based data augmentation approach which was based on pitch shifting.…”

Multi-task learning for arousal and sleep stage detection using fully convolutional networks

“…It enhances the important parts of its input and fades out the rest, so that the network can focus more on the relevant information and less on the noise [71]. In this work, we use the attention mechanism reported in [72] with 256 units. The attention mechanism calculates the context in three steps.…”

CovidCoughNet: A new method based on convolutional neural networks and deep feature extraction using pitch-shifting data augmentation for covid-19 detection from cough, breath, and voice signals

“…Delayed test results accelerate the spread of the virus. The development and implementation of rapid, scalable, reliable, economical, and reproducible tools for diagnosing and detecting Covid-19 are extremely important [ 10 , 11 ].…”

“…Different respiratory diseases such as pertussis, asthma, pulmonary edema, tuberculosis, Parkinson's, and pneumonia have been detected using AI methods based on cough or sound signals [ [21] , [22] , [23] , [24] , [25] , [26] ]. In addition, AI techniques have been developed to detect Covid-19 by analyzing cough, breath, and sound data [ 4 , 6 , 7 , 10 , 11 , 19 , [27] , [28] , [29] , [30] , [31] , [32] , [33] , [34] ].…”