An Open-Set Recognition and Few-Shot Learning Dataset for Audio Event Classification in Domestic Environments

The problem of training with a small set positive samples is known as few-shot learning (FSL). It widely that traditional deep algorithms usually show very good performance when trained large datasets. However, in many applications, it not possible to obtain such high number samples. This paper deals the application FSL detection specific and intentional acoustic events given by different types sound alarms, door bells or fire using limited These sounds typically occur domestic environments where corresponding wide variety classes take place. Therefore, alarms practical scenario can be considered an open-set recognition (OSR) problem. To address lack dedicated public dataset for audio FSL, researchers make modifications on other available aimed at providing community carefully annotated dataset1 OSR context comprised 1360 clips from 34 divided into pattern unwanted sounds. facilitate promote research this area, results state-of-the-art baseline systems based transfer are also presented.