We need your help. For several years scientists have been trying to extract meaningful information about music directly from the audio data in a track. Encouraged by successes in automatic speech recognition, for a while researchers hoped that it would be fairly easy for a computer to analyze an audio track and transcribe it directly to sheet music notation, recognising instruments, voices, and all the notes they were playing.
This was quickly found to be a much harder problem than expected, and a new field of research grew out of the failure to conquer this challenge.
Since then scientists have battled to recognise specific aspects of a song, such as its tempo, key, harmony, rhythm, instrumentation, or even just its genre, by analyzing its audio content. Starting in 2004, this battle has been formalised into an annual competition called MIREX.
In each round of MIREX, competitors submit their programs to a carefully controlled competition server, where each program is run on the same set of tracks. The winner is the one whose results agree most closely with a matching set of human judgements.
But gathering human opinions about very specific aspects of large numbers of songs isn’t easy. You need to find people with a keen interest in music, and ideally some musical expertise. You need to persuade them to answer questions that can seem obvious to them (even though they are still difficult for a computer). Finally you need to have enough helpers to allow for the fact that some interesting musical questions can have more than one ‘right’ answer.
We’ve been thinking about this problem recently, while evaluating some rival algorithms to estimate the tempo, or beats per minute, of songs directly from audio… and we thought of you! We know from many years’ experience that Last.fm users have the interest, the knowledge, and the sheer staying-power to make a huge contribution to scientific research in this field. As an extra incentive, improving these methods could help us offer you even better radio and recommendations in the long run.
To test this idea in practice, we’ve built a very simple application on our labs website where you can listen to music and help us improve the state of the art in tempo estimation. If you have fun with this, we plan to add to it over the coming weeks and months.
So please lend us your ears for a few minutes. Thank you!
Comments
Tecfan
12 May, 10:50
Fun stuff! Can’t believe I’m doing this without getting scrobbles :P
Anyway, found two bugs:
Usernames are hidden in the leaderboard (in FF4 for me).
If the song changes when select speed, and you want to go back to the first song to tap, the tap-o-meter does not appear. (does this make any sense?)
Tecfan
12 May, 10:53
Also, if you change opinion about the speed after you have tapped the tempo (a song I got started very slow the first 3 seconds, so I just pressed slow right away), the “TAP THE SPACEBAR” text goes over your result as if there is no result.
Baba
12 May, 11:47
“even just it’s genre”?
Shouldn’t play that down. You’ll find that much harder than those other things if you want to be exact with your genres.
fuzy83
12 May, 11:53
it would be interesting to see what others answer
Niviel
12 May, 12:10
Lovely idea, and quite fun too :)
Tecfan
12 May, 13:17
http://tecfan.nitrolinken.net/s/lastfmspeedo.png \o/
Axefield
12 May, 13:49
Maybe you can have like wiki user submitted lyrics, that then can be interpreted by some sort of learning AI
epix
12 May, 14:31
gr8! im almost hoping that someday i’ll work for u :) keep UP good work
daniel
12 May, 14:48
Nice exercise.
wesley
12 May, 15:07
I feel dumb, but I can’t figure out what the main beat is supposed to be. Is it snare, or snare + kick drum, or something else?
Andrew Clegg
12 May, 16:25
Hey Wesley, think of it as “the bit you tap you foot to”. What that actually is varies from genre to genre. Rock might go bass-snare-bass-snare, trance might go bass-bass-bass-bass, etc…
Sam, defend your premise at
12 May, 17:11
I’m straight-up amazed that (whoever) has funded this.
I can only imagine afew assumptions and the seem silly. The real ones no doubt, have 97 Playstations strung together (DARPA’s “worlds fastest supercomputer”) hung.
Are we assuming that every instrument in the acoustically perfectly dead room is perfectly tuned to it’s proper spot against (whichever) concert A?
Are we assuming that, oh say chair 1 violin is, indeed a robot and can play with techical perfection and pull off “no-feel?” or,
Are we assuming that a totally untrained musical “monkey banging out shakespeare on 1/10e6 of 10e6 typewriters with monkeys,” (such as myself) got a stroke of genius in Reason 4.0 and came out with a listenable tune one night at 04:67h?
This one gets a resounding C’mon Y’all.
Jon Smirl
12 May, 21:57
Totally unrelated but I’d love! a way to disable Christmas music via my profile settings and then turn it back on in December.
I am using the mix channel and I am getting tired of banning hundred and hundred of Christmas songs.
henriqueqc
12 May, 23:20
I’d like to see that turned into a game. Maybe using some ideas from Luis von Ahn. Altought I’m having fun with it anyway! Great work!
Pichu0102
13 May, 03:29
On Chrome, I don’t see any sort of bpm indicator, just a “Tap space bar to measure bpm.
Jon Hallier
13 May, 10:05
Jon, we have this feature. :) Go to your profile settings (http://www.last.fm/settings/website), and enable Bah Humbug! mode. Previously there was a problem with it not working with mix radio, but that should be fixed now.
E-Clect-Eddy
14 May, 18:55
Looks like it does not work properly in Opera 11.1 – XP SP3, or I just don’t know how to get to the next song after recording of first track.
Herman
17 May, 17:05
Is it down? It won’t load in Firefox 4.0.
Erdi
21 May, 16:00
Hi.i am working on speech recognition at msc.i had developed simple data mining tool about lastfm using lastfm api at undergraduate.now,i visit on mobile phone.i will try to visit again for all of details of your project.i look for thesis subject.i can be interesting your project.
Sincerely,
erdi
Cat
21 May, 18:48
very cool. hope it works out! :]
jer2112
22 May, 05:57
i wonder how plant growth would respond to it.?
Mike R
22 May, 14:15
For the second song it says recording but it never indicates when its done recording. So I end up click save my answers. Also after the third page of songs it won’t let me listen to the songs to determine whether the tempo is slow,in between or fast. Please fix these problems.
C26000
23 May, 06:47
fun dancing by myself to feel better the beat ;)
javier vallejos(fifty still)
24 May, 23:53
dime como querer mas la musica que ago. tocando o generando palpito. lo escucho lo canto en todo momento en la vo en silencio… solo dy que el padre del rap en argentina habla. fifty still paz
black_napkin
25 May, 19:35
wait wait wait. I’ve tried a few, and I think there’s a problem with this setup. After doing a few, you’ll come to expect that the songs will have the same tempo but “feel” faster or slower. And even though the second song might seem to have a different speed, I can confirm to myself that they are the same speed by remembering how fast I tapped for the first song.
So should I go with my immediate feeling (that they are different speeds) or what I know after thinking about it (they are the same speed)?
Hopefully this made some sense.
bean
25 May, 22:23
I’m using Chrome, and I have to hit play a few times to actually get it to play
BenK
25 May, 23:38
If you really want to extract information from songs, it will probably have to be an iterative, multi-pass analysis.
JeremyR
27 May, 07:14
Maybe Last FMs scientists can figure out a way of separating different artists with the same name…
neoniko
27 May, 09:06
one guy go to a microfone and go sing-oh my love ,,,
ashmonkeyuk
27 May, 20:20
I’ve been looking into this for a long while as well, looking at it from a philosophy of semantics point of view. I’m pleased to take part and would like to get involved more!
Igor
31 May, 08:39
Cool stuff! But I have found one thing – my results seem to be skewed when, for example, one space bar hit results in a repeating key event. It makes the BPM reading go skyrocketing to 40000, and later it never comes back to the actual reading.
Maybe you should somehow track these things, and allow only a certain amount of taps per, say, 30 milliseconds?
Some 50% of the app runs that I have completed so far have been affected one way or the other by a problem of this nature – one wrong or missed space bar beat can totally ruin the whole reading. My results were offset this way for 20 or 30%.
I hope I make some sense :)
Radio F.C.
1 June, 14:08
This is interesting. I would like to see you display the average BPM when available for the tracks I have scrobbled or better still any tracks anyone else scrobbled. Perhaps you could do this like the images for artists is done currently i.e. let users enter a value and have others vote on the accuracy. This would be VERY VERY USEFUL for DJs preparing playlists… Thanks.
waldemar waldecir
1 June, 22:23
sou fa das musicas anos 50 me criei ouvindo elas
Tlanextik
1 June, 22:51
This is interesting indeed. Good for science and good for music.
I have a question: what does ‘worth x points’ means?
cOOp
3 June, 01:55
hasn’t echo nest already done this ?http://visualizer.fm/static/index.html#
Mick Voo
7 June, 00:51
Filter into about 10,000 discreet frequency bands and build profile interaction models. I’ll settle for 45%.
vathorniel
8 June, 10:04
I never actually thought about it but it’s so weird when you start doing this speedo thing and the second song seems CLEARLY slower \ faster, until you start tapping and find out that they’ve absolutely the same beat rate :-)
KC
10 June, 18:09
Aliens have already accomplished this, 74,329 years ago (I may be wrong on the exact amount of years) ;-)
Xel
11 June, 03:23
This is fun! I think this would be a great addition with some improvements. It also has introduced me to new music! You give a song more of a chance when you’re listening for a beat.
aceng
14 June, 22:44
this is so great it makes people can sharing with each other
Uncledody
15 June, 00:58
music changes with one’s mood…it’s a state of mind at the moment…hard to quantify.
shobhnaguerin
17 June, 06:41
This is a great initiative!
fetzu
18 June, 15:18
Cool stuff. I’ll try to do small 10/15 minutes sessions from time to time.
Hope this helps improve the state of so called “bpm indicators”.. because clearly huge amount of improvement can be made in that area!
brokenblog
18 June, 20:54
i failed the task. real masterpieces live and develop, let us believe a special rythm, but lie and persuedes us to fallow another rythm, plays with our emotions, bluffs us, hide the rhythm, later reveals the truth, let us realize when its too late, so its very important to listen to the whole track. So i am sorry, i stopped the test.
tristanllg
22 June, 05:00
I’m pleased to take part and would like to get involved more!
Comments are closed for this entry.