Proposed Experimental Design of a Portable COVID-19 Screening Device Using Cough Audio Samples

With the proliferation of COVID-19 cases, it has become indispensable to conceive of innovative solutions to abate the mortality count due to the pandemic. With a steep rise in daily cases, it is a known fact that the current testing capacity is a major hindrance in providing the right healthcare for the individuals. The common methods of detection include swab tests, blood test results, CT scan images, and using cough sounds paired with AI. The unavailability of data for the application of deep learning techniques has proved to be a major issue in the development of deep learning-enabled solutions. In this work, a novel solution of a screening device that is capable of collecting audio samples and utilizing deep learning techniques to predict the probability of an individual to be diagnosed with COVID-19 is proposed. The model is trained on public datasets, which is to be manually examined and processed. Audio features are extracted to create a dataset for the model which will be developed using the TensorFlow framework. The trained model is deployed on an ARM CortexM4 based nRF52840 microcontroller using the lite version of the model. The in-built PDM-based microphone is to be used to capture the audio samples. The captured audio sample will be used as an input for the model for screening. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Hybrid CNN-LSTM Speaker Identification Framework for Evaluating the Impact of Face Masks

Following the declaration of COVID-19 as a worldwide pandemic, hindering a multitude number of lives, face mask exploitation has become extremely crucial to barricade the emanation of the virus. The masks available in the market are of various sorts and materials and tend to affect the speaker's vocal characteristics. As a result, optimum communication may be hampered. In the proposed framework, a speaker identification model has been employed. Also, the speech corpus has been captured. Then, the spectrograms were obtained and passed through a two-stage pre-processing. The first stage includes the audio samples. In contrast, the second stage has targeted the spectrograms. Afterward, the generated spectrograms were passed into a hybrid Convolutional Neural Network- Long Short-Term Memory (CNN-LSTM) model to perform the classification. Our proposed framework has shown its capability to identify speakers while they are wearing face masks. Moreover, the system has been evaluated on the collected dataset, where it has attained 92.7%, 92.62%, 87.71%, and 88.26% in terms of accuracy, precision, recall, and F1-score, respectively. The acquired findings are still preliminary and will be refined further in the future by data expansion and the employment of numerous optimization approaches. © 2022 IEEE.