EARSAP

we manually listened to all the sounds and classified them into four categories: [Type-0] Sounds that are completely irrelevant to the corresponding labels; [Type-1] Earcon (The identification of Earcon is based on the guideline of Earcons design reported by Blattner, Sumikawa and Greenberg); [Type-2] Auditory icon (which is made by capturing the sounds in everyday life); [Type-3] Music/vocal/other sounds that are relevant to the label description. To study the relationship between semantic descriptors and sound features of Earcon, we first obtained 270 earcons with short, structured musical messages from a mixed database containing 1053 sound samples with different types.