Modern music production systems provide a plethora of sound resources, e.g. hundreds or thousands of sound patches on a synthesizer. The more the number of available sounds grows, the more difficult it becomes for the user to find the desired sound resource for a particular purpose, thus demanding for advanced retrieval techniques based on sound classification. This paper gives a short survey of existing approaches on classification and retrieval of sound resources, discusses them and presents an advanced approach based on ontological knowledge processing.