Can you automatically recognize sounds from a wide range of real-world environments?