About Database in pseudo prompt enhancement

Thanks for great work!
Could you please confirm if the database is directly sourced from the audio sources mentioned in the paper without any processing, or if there has been some processing, such as filtering out samples containing only single sound events?
