HAASD: A dataset of Household Appliances Abnormal Sound Detection

Intelligent household appliance sound event detection and classification is an evolving research field for intelligent diagnosis and evaluation of household appliances. In this paper, we identified three major barriers to research in this area-the lack of a common taxonomy, the scarcity of negative samples, and the low signal-to-noise ratio of household appliances' sound signals. In order to solve these problems, we proposed appliance fault or abnormal sound detection and a new dataset household appliances abnormal sound detection (HAASD), which is divided into two categories: normal sound and abnormal sound. Each category has more than one background noise file. Noise data annotated in the mode. A series of experiments using the baseline classification system were used to study the challenges of the data set, and multiple evaluation indicators of different characteristics in different classifiers were compared.