Win a copy of The Little Book of Impediments (e-book only) this week in the Agile and Other Processes forum!
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic

hwo to retrieve an audio from a databse by a hummed tune

 
mags halligan
Greenhorn
Posts: 2
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi,
im trying to create an application for a project that allows the user to search for a tune in a databse by humming it.

It currently captures the hummed tune and stores it in an audioinputstream. I'm not sure how to compare the hummed file with the wav file in the database to retrieve it. Do i have to do an FFt on the hummed tune then a spectrum analysis. Any ideas or suggestions would be a great help.

Thanks in advance
Mags
 
Steven Bell
Ranch Hand
Posts: 1071
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
I seem to remember this being tried before somewhere. It would seem to me that it would be just as difficult, or more, as voice recognition (especially considering some people are completely tone deaf )

Some questions that might help:

Does the user need to hum the whole song or just a portion? what portion? what is the minimum allowed (I can name that tune in 2 notes )?

How far off pitch/tempo can the user be?

How do you account for tone quality or sound type? (I'm not a sound expert, but I know a piano sounds different than a hum)

Is there a way you can index the music by factors such as tempo/type?

Would having addition input help, such as the user selecting a type of music to limit the search?

Interesting project though. Good Luck.
 
mags halligan
Greenhorn
Posts: 2
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
hey thanks for that. it doesnt have to be complex at all. just a portion of the tune hummed and then compared with the databse. it just has to be a very a simple prototype, maybe kinda of comparing the two files. would you have any suggestions on how to do a real simple version i havent a clue
 
Steven Bell
Ranch Hand
Posts: 1071
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Like I said I'm not a sound expert, but I would probably start by having a piece of music and a hum that should match that music. Then look at the two pieces, and look for similarities (I would hate to think you would need to look at the bytes, but it may come down to that). Maybe you could look for beats and just do a rythem comparison (distance between beats, beat patterns). I know visually beats are pretty easy to spot, not sure how hard it is to find them programatically.
 
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic