Microsoft’s online chief holds music patent
When Microsoft employed Qi Lu to run its internet business last week, the company trumpeted the indisputable fact that Lu holds twenty patents.
Patents are a long way from rare at Microsoft–many developers and analysts hold them–but the web business has usually been controlled by folk with a business or selling background.
That has not been working out too well, so it’s putting a geek in charge. Most interesting to me, one of them is linked to music. Particularly , it describes a Computer application that would take a snippet of a song or audio file, break it down into element parts, research them, and then counsel similar songs.
It sounds superficially like what Shazam does, but the strategy is absolutely different and more sophisticated.
From what I will tell, Shazam simply takes a sound sample and matches it against a database with millions of audio files. Getting a fast result requires some fast info crunching, but there’s not too much deep research going on there. Lu’s patent ( shared with 2 other engineers ) suggested breaking the song all of the way down to miniscule parts like measures and individual notes, researching those elements to find patterns–for example, a repeated sequence of notes could be the desist or chorus–and then researching the relations among those parts. As an example, a pop song is usually built of many repeated verses and choruses, with a bridge somewhere in the middle. This is the way in which the application would be ready to identify and recommend songs that are like the tune being played. Rather than Shazam, the final result could have been more like Apple’s lately introduced Genius feature, which builds playlists of songs based primarily on the track you are now playing. I think that Apple’s depending on information from all its iTunes users ( Genius asks to gather info about your playing habits ) and song meta data–for example, it frequently endorses songs by the same artist, or other artists in the same class, or other songs released in the same era. That is much easier–both to program and for your CPU–than attempting to research audio info for patterns.
Lu received this patent in two thousand, meaning that he was potentially working on it many years before that.