“…However, the previous challenge was mitigated in 2019 with the emergence of the largest public database, the International Conference on Biomedical and Health Informatics (ICBHI) [ 60 , 61 ]. Therefore, research focused on different machine learning approaches has recently increased dramatically, such as Recurrent Neural Networks (RNN) [ 62 ], hybrid neural networks [ 63 , 64 , 65 , 66 , 67 ] and above all Convolutional Neural Networks (CNN) [ 64 , 68 , 69 , 70 , 71 , 72 , 73 , 74 , 75 , 76 , 77 , 78 , 79 , 80 , 81 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 , 90 , 91 , 92 , 93 , 94 , 95 , 96 , 97 , 98 , 99 ]. Thus, the use of these types of deep learning architectures provided promising performance improvements due to their ability that they are able to learn behaviour, both in time and frequency, from large datasets, eliminating the engineer intervention in feature extraction techniques, which reduces the likelihood of human error [ 100 ].…”