|
|
| Author |
Message |
| Redoko |
This post is not being displayed .
|
 Redoko World Chat Champion

Joined: 04 Nov 2009 Karma :    
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| Redoko |
This post is not being displayed .
|
 Redoko World Chat Champion

Joined: 04 Nov 2009 Karma :    
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| soforene |
This post is not being displayed because the poster has bad karma. Unhide this post / all posts.
|
 soforene Nearly there...
Joined: 25 Oct 2009 Karma :   
|
 Posted: 13:52 - 31 Jan 2012 Post subject: |
 |
|
Surely you mean Left Bank Two by The Noveltones.
I thought everybody knew that !!  |
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| daemonoid |
This post is not being displayed .
|
 daemonoid World Chat Champion

Joined: 27 Jun 2008 Karma :    
|
 Posted: 14:21 - 31 Jan 2012 Post subject: |
 |
|
There's an app called sound hound - I got it for free with apple's 12 days of Christmas. It's similar to shazam in that it can recognise songs on the radio etc. but you can also sing at it and it's surprisingly effective - https://www.soundhound.com/
As a software engineer I find both of these apps astounding - I wouldn't really know where to start in developing something similar that returns results as quickly as they do. Really impressive tech. ____________________ current: ducati monster 750
past: hyosung gt250r, bajaj pulsar 180, hyosung gt 125 comet
@thomasgarrard | www.straitjkt.com | www.racingseven.com |
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| Frost |
This post is not being displayed .
|
 Frost World Chat Champion

Joined: 26 May 2004 Karma :  
|
 Posted: 18:34 - 31 Jan 2012 Post subject: |
 |
|
There is a technique called Fourier Analysis:
https://en.wikipedia.org/wiki/Fourier_analysis
Which basically states that any noise can be made up out of a collection of other frequencies. This can be applied to music, images or even video! Anyway this can be used to break a piece of music down into a series of parameters that can then be searched for on a database. For instance a hashing function could be applied to the parameters and deliver the title of the piece from a list without performing any form of search at all. That would be very very fast indeed but would require an exact match. AI techniques could be used to look for a best match however and would be far more robust.
I'd be tempted to make something like it except for:
I should probably be spending my time doing something more productive like getting a job.
There are already so many out there that mine would go totally unnoticed. |
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| daemonoid |
This post is not being displayed .
|
 daemonoid World Chat Champion

Joined: 27 Jun 2008 Karma :    
|
 Posted: 20:47 - 31 Jan 2012 Post subject: |
 |
|
I get that, but then how do you search any 4s sample against a 'database' of such a vast collection of music?
and then sound hound with it's ability to detect by humming or singing...
I'm hard to impressed with tech, but the more you think about this problem the harder it gets so these solutions really astound me. ____________________ current: ducati monster 750
past: hyosung gt250r, bajaj pulsar 180, hyosung gt 125 comet
@thomasgarrard | www.straitjkt.com | www.racingseven.com |
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| Phoenix |
This post is not being displayed .
|
 Phoenix Twisted Firestarter

Joined: 01 Aug 2002 Karma :    
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| The Artist |
This post is not being displayed .
|
 The Artist Super Spammer

Joined: 06 Jan 2008 Karma :  
|
 Posted: 21:49 - 31 Jan 2012 Post subject: |
 |
|
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| angryjonny |
This post is not being displayed .
|
 angryjonny World Chat Champion

Joined: 01 Sep 2006 Karma :    
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| Frost |
This post is not being displayed .
|
 Frost World Chat Champion

Joined: 26 May 2004 Karma :  
|
 Posted: 01:30 - 01 Feb 2012 Post subject: |
 |
|
| daemonoid wrote: |
I get that, but then how do you search any 4s sample against a 'database' of such a vast collection of music? |
You can use something like a tree with a heuristic, that is a measure of similarity between the current sample and another sample. So it could for instance compare the frequency of the bass to other samples and go:
Slower than this, slower than that by a bit, faster than that, so it's a song from this sub tree. Straight away that's chopped it down from one of 100,000 songs to one of say 5000. Then finer matching can be done on other frequencies.
A mate did something similar with some voice control software that was trained from some samples you provided, it then used short term energy readings to segment the signals, then DCT and DFT to break it down into some parameters that could be used to match samples. I believe he used Naive basien and Gaussian classifiers both to good effect. He got a few good papers published on it. |
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| fatpies |
This post is not being displayed .
|
 fatpies World Chat Champion

Joined: 01 Mar 2011 Karma :   
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| The Shaggy D.A. |
This post is not being displayed .
|
 The Shaggy D.A. Super Spammer

Joined: 12 Sep 2008 Karma :  
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
| fatpies |
This post is not being displayed .
|
 fatpies World Chat Champion

Joined: 01 Mar 2011 Karma :   
|
|
| Back to top |
|
You must be logged in to rate posts |
|
 |
Old Thread Alert!
The last post was made 14 years, 237 days ago. Instead of replying here, would creating a new thread be more useful? |
 |
|
|