Navigation | ||
My WhyNot? | ||
Icon Key | ||||||||||||
The community's level of support for this idea is...
| ||||||||||||
Google Tune | |||||||||||||||||
A means of looking up tunes on the Internet in order to identify them. Who hasn't had a tune stuck in their head, and gone around for days, wondering, "What the heck is this song?!" Not sure how to implement it (maybe searchers could use some kind of midi interface and enter a tune fragment by playing it). Then the application would have to run a sort of similarity search. I have sunk low enough to call up DJ's and sing my fragment to them; it's embarassing.
schaplan, Nov 12 2006
What do you think of this idea or comment?
| |||||||||||||||||
I agree
No opinion
I disagreeUsers who liked this idea also liked: | ||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||
Add your comment
A company could compile every known digital musical file, including midi.
A set of digital signal processors could then pull out lets say 50 characteristics like lyrics, sequence of tones, number of occurences of each tone, phonetic qualities, relative pause durations between parts, ... and generate a profile of characteristics.
Each music file would then be associated with each of the data sets. Indexes would be created and sorted based upon the different characteristics.
An intelligent feedback system would raise the weighted values of characteristics that commonly have greater success in determining the tune of interest for users.
The user would navigate the web to that company's website and download an active-x component into their web browser. The active-x component would ask the user to use a microphone and hum or sing the tune into a computer microphone.
The active-x component would package the mp3 file and send it to the company. The company computer would then run the mp3 file through the same set of digital signal processors and generate the profile of characteristics. The indexes would be compared with the submitted profile of characteristics and close matches noted (fuzzy logic).
The compared matches would identify perhaps 850 different songs that had some possibility of matching. The songs with higher correlation would be listed first. The user would page through the songs until they found the one they were looking for. Upon selecting the song as "Found", the company computer increases the statistical weighting of each of the digital signal processor outputs that were related to identifying the tune. In this way, the ranking will change and the next time a user hums a song, the actual song will appear higher in the ranking.
Other intelligent selection algorithms are feasible, including neural nets for adjusting DSP output selection criteria.
Rogers Wireless (Canadian mobile phone provider) offered such a service before.