Wavelet-Based Cough Signal Decomposition for Multimodal Classification

Abstract

Signal classifications have benefited from the successes of ML and DNN architectures. Cough classification techniques mainly extract features such as the Mel Frequency CepstralCoefficients for training. Most of these works also focus on obtaining information from single data modalities. However, multimodal analysis has been shown to aggregate useful information from different modalities thereby improving the internal capacity of ML models at data analysis. In this research, we propose a multimodal cough data classification approach with scalograms images obtained by decomposing cough signals using continuous wavelet transform and clinical information of subjects obtained from the COUGHVID dataset. Our result shows improved precision as compared to expert analysis.

Multi-modal Point-of-Care Diagnostics for COVID-19 Based On Acoustics and Symptoms

“…On a similar dataset, Coppock et al [22] report 0.85 AUC. Agbley et al [23] demonstrated a specificity of 0.81 at a sensitivity of 0.43 on a subset of COUGHVID dataset. Feng et al [24] used a subset of cough sounds from Coswara dataset and reported a performance of 0.90 AUC.…”

“…Laguarte et al [20] obtained AUC greater than 0.90 on samples from the COVID-19 Cough data set. These studies use acoustic feature representations of cough sounds such as Mel frequency cepstral coefficients (MFCCs) [18], Mel-spectrogram [20], [22], or scalograms [23], while the classifier models are deep learning based neural networks such as convolutional neural networks (CNNs) [23], recurrent neural networks (RNNs) [24], CNN based feature embeddings in support vector machines (SVM) [18] or with CNN based residual networks [20], [22]. There are also attempts at creating more controlled COVID-19 cough sound dataset from individuals in hospitals [25], [26].…”

A Multi-Branch Deep Learning Network for Automated Detection of COVID-19

“…We adopt a three class diagnostic system previously created by Agbley et. al [2] -COVID-19 positive, symptomatic COVID-19 negative, and asymptomatic COVID-19 negative. Furthermore, by adding novel model features, we show that our model can better distinguish between COVID-19 positive patients and COVID-19 negative patients than any previous approaches, regardless of symptom status.…”

“…In table 4.3, we compare the performance of our network in detecting COVID-19 to another deep learning model [2] trained on the COUGHVID dataset, along with medical experts' prediction performances. Our network outperforms both experts and Agbley et al [2] in detecting COVID-19 from cough audio files.…”

“…The state-of-the-art performance of AI algorithms for detecting COVID-19 ranges from 0.68 AUC to 0.97 AUC [2,15], based on several factors, such as the quality of audio files used to train the learning algorithm, preprocessing methodology, and the algorithm structure. In our previous work [6], our group developed a deep learning model with an average AUC of 0.77 across several datasets, including Coswara [19], COUGHVID [16], and Virufy [6] datasets.…”