Skip to main content

Professor Les Atlas (PI)
Ruobai Wang, graduate student in electrical engineering will allow users to discover new music through music they already love. Using a super-quick match approach called “context sensitive hashing,” along with the power of machine learning based signal processing, is different from usual content-based music retrieval systems in that it matches segments of music instead of whole songs.

An initial mock-up of the consumer interface for simsong.

The easy-to-use software functions allow users to quickly locate a new segment of music, which is similar to a segment that the user already likes. For example, classical music segments have names such as passages, and rock has segments called riffs or hooks. The software will capture the unique qualities of these sounds and find similar passages in in other, perhaps unheard, songs. This is a new approach to find new and interesting music, kind of like automatically “digging in the crates” in a traditional music store, but on a much larger and world-wide scale.

University of Washington (UW) electrical engineering graduate student Ruobai Wang is working with Professor Atlas to put together this browser-based application, which uses the cloud to match a favorite portion of a song to essentially all music in a large catalog.

For Wang, simsong opens up new discoveries of untapped music. In his examination of the software, he downloads piano music from one of his favorite YouTube videos and “simsongs” his favorite passage within it. One of the three best matching music segments simsong finds astounds him: a beautiful, eclectic song from an obscure band from Finland.

In addition to finding new music, simsong also removes barriers to foster a new global stage for music discovery. Artists from all over the world can be accessed, not only opening up parts of the world to new music, but also allowing the music industry to become more multi-national.


Research Areas