Speaker: Miguel Ángel. Abstract: Analysing respiratory audios in order to detect and classify adventitious respiratory sounds is of vital importance for the development of continuous monitoring tools for patients with respiratory diseases. The ICBHI 2017 database is the most widely used database in this field of research, but its small size is a major limitation for the complexity of the models to be employed. Therefore, a novel data augmentation technique is presented where new artificial respiratory cycles are created using Harmoic-Percussive Decomposition.
