A New COVID-19 Detection Method Based on CSK/QAM Visible Light Communication and Machine Learning

This article proposes a novel method for detecting coronavirus disease 2019 (COVID-19) in an underground channel using visible light communication (VLC) and machine learning (ML). We present mathematical models of COVID-19 Deoxyribose Nucleic Acid (DNA) gene transfer in regular square constellations using a CSK/QAM-based VLC system. ML algorithms are used to classify the bands present in each electrophoresis sample according to whether the band corresponds to a positive, negative, or ladder sample during the search for the optimal model. Complexity studies reveal that the square constellation N = 2(2i )x 2(2i), (i = 3) yields a greater profit. Performance studies indicate that, for BER = 10(-3), there are gains of -10 [dB], -3 [dB], 3 [dB], and 5 [dB] for N = 2(2i) x 2(2i), (i = 0,1, 2, 3), respectively. Based on a total of 630 COVID-19 samples, the best model is shown to be XGBoots, which demonstrated an accuracy of 96.03%, greater than that of the other models, and a recall of 99% for positive values.

Automated summary

This article proposes a novel method for detecting COVID-19 in an underground channel using VLC and ML. The mathematical models of COVID-19 DNA gene transfer in regular square constellations using a CSK/QAM-based VLC system are presented. ML algorithms are utilized to classify the bands in electrophoresis samples to search for the optimal model. Results show that the square constellation N = 2(2i )x 2(2i), (i = 3) yields the highest profit and XGBoost achieves the best accuracy of 96.03% and a recall of 99% for positive values among all the models tested.