The Amazing Music Identification Technology

[audreyh1]

Have any of you tried the various methods of "name this tune"? We've used it a couple of times to try to remember the name of a song, asking Siri to listen and identify.

We were watching an episode of Poirot (had Tim Curry as one of the leads, no less, LOL!) and they played a lovely baroque vocal bit that sounded vaguely familiar. Realizing the words were in English, I knew it was one of the English baroque composers - maybe Handel or someone. So I put Siri to the test, she listened, and after several seconds comes back with:

Famous Baroque Arias
Robert King & The King's Consort
Dido & Aeneas: When I am laid in Earth (Henry Purcell)

Yep - that was it, bingo. Not only the piece of music, but the exact recording - with a link to buy it, of course!

This technology just blows me away! There was even some dialog during the music on the show - didn't matter.

How many 10s of thousands of hours (if not 100s of thousands) must there be of recorded music, including oodles of classical music? I just don't understand how the searches can be so fast.

 

[ERD50]

From what I understated, it only works with recorded tracks, matching the recording it 'hears' to its library. A 'fingerprint' is created, based on certain parameters of that music (the beat? relative levels of different frequencies changing over time? a bunch of other things?). I'm sure there is some super advanced modeling and pattern recognition stuff going on, that only people with a PhD in these areas can even begin to grasp.

What I really can't understand is, how it can match a few seconds from a random starting section of a tune, to anything, at any point, in its library. Wouldn't you need to take that fingerprint, and 'slide' it through every song, start to finish, to find the match? Wouldn't that be thousands, millions, of comparisons for each tune it needs to check until it finds a match? Maybe they can eliminate a bunch of songs that don't have that kind of signature anywhere, but it still boggles my mind.

[off topic rant]And if software can be this sophisticated, and available to the common man, why can't it be super easy to find which doctors are in my network for the various health care plans I'm being offered? Seems I need to sort through, with a unique search for each plan and each doctor. That's a much smaller universe.[/off topic rant]

Conspiracy theory: They have been implanting subliminal 'watermarks' in all songs for decades. We can't hear them, but every 1/2 second, an identifying 'bar code' is encoded, to identify the song. All done for copyright protection purposes.

Not-so-Conspiracy theory: As I understand it, this song identification technology was developed to protect copyrights, and that is why youtube will challenge various uploads, even those with incidental music in the background.

-ERD50

 

[audreyh1]

What I really can't understand is, how it can match a few seconds from a random starting section of a tune, to anything, at any point, in its library. Wouldn't you need to take that fingerprint, and 'slide' it through every song, start to finish, to find the match? Wouldn't that be thousands, millions, of comparisons for each tune it needs to check until it finds a match? Maybe they can eliminate a bunch of songs that don't have that kind of signature anywhere, but it still boggles my mind.

-ERD50

Yeah - that's what has me scratching my head

 

[tulak]

SoundHound app on my iPhone.

I don't know if it would have worked in your case, but so far every time I've used it it's worked.

 

[Midpack]

This technology just blows me away! There was even some dialog during the music on the show - didn't matter.

How many 10s of thousands of hours (if not 100s of thousands) must there be of recorded music, including oodles of classical music? I just don't understand how the searches can be so fast.

Wow, I am truly amazed!

I went out of my way to pick less well known songs & artists. Siri got tunes & artists exactly correct by American HiFi, Social Distortion, Sixpence None The Richer, Vince Guaraldi Trio, Todd Rundgren, Marc Cohn, Shania Twain, Shawn Colvin, Randy Newman, Maxi Priest and from Rocky Horror Picture Show and Les Mis.

However, she may know popular music genres better than others.

She did not recognize John Williams original Jurassic Park theme (surprising IMO, broadly known) at all and guessed way wrong on a classical guitar Partita in E Major by Ana Vidovic (an almost unfair choice, arguably very obscure).

 

[tmm99]

SoundHound app on my iPhone.

I don't know if it would have worked in your case, but so far every time I've used it it's worked.

I use SoundHound too. It's a great app IMHO.

You got to try humming into the app and see if it can identify which song you are humming! If it can, you are certifiably good at humming IMO. I even hummed a Japanese song from the 80's and strangely enough, the app played back someone else humming the same song (The real song must not have been loaded yet - When I try it now, it shows the song from iTune).

You got to try it out - It's free.

 

[NW-Bound]

Does it work if you sing or hum out of tune, as I am apt to do?

 

[tmm99]

Does it work if you sing or hum out of tune, as I am apt to do?

It may do word recognition, I don't know. (Try it out and see.) I'm pretty sure though, it won't be able to identify if you hum badly.

 

[NW-Bound]

...
[off topic rant]And if software can be this sophisticated, and available to the common man, why can't it be super easy to find which doctors are in my network for the various health care plans I'm being offered? Seems I need to sort through, with a unique search for each plan and each doctor. That's a much smaller universe.[/off topic rant]

Heck, what you want is way too advanced for them.

I have told the story earlier of how my health insurer did not apply the payment I made online to my account. One of their Web pages showed that I made the payment, yet another of their pages showed an overdue balance. And I received multiple letters from them threatening to cut off my policy. It took multiple phone calls (4 or 5) before they could correct it.

For 2016, I kept this insurer, but changed from a Silver plan to a Bronze. Again, one of their Web pages acknowledges the new policy, but the billing page shows a balance of 0. There's no way for me to pay, so I had to call them. And now, it shows that I have some credit because the premium is still not billed.

I guess all the good programmers go to work on smartphone apps, and leave many fields to incompetent programmers.

 

[Nodak]

Another Soundhound fan here.

 

[marko]

It may do word recognition, I don't know. (Try it out and see.) I'm pretty sure though, it won't be able to identify if you hum badly.

I use Shazam which is similar. IIRC it identifies a profile of music/tonals and time against known recorded library. Playing the song on your kazoo won't work.

It is the same technology that spawned Pandora (sort of in reverse) where you tag songs that you like and it learns to send you similar songs based not on the artist but on the type of music (that is why if you tell Pandora that you like Jimmy Buffet, you get a lot of Country music instead of island music).

 

[calmloki]

Have any of you tried the various methods of "name this tune"? We've used it a couple of times to try to remember the name of a song, asking Siri to listen and identify.

We were watching an episode of Poirot (had Tim Curry as one of the leads, no less, LOL!) and they played a lovely baroque vocal bit that sounded vaguely familiar. Realizing the words were in English, I knew it was one of the English baroque composers - maybe Handel or someone. So I put Siri to the test, she listened, and after several seconds comes back with:

Famous Baroque Arias
Robert King & The King's Consort
Dido & Aeneas: When I am laid in Earth (Henry Purcell)

Yep - that was it, bingo. Not only the piece of music, but the exact recording - with a link to buy it, of course!

This technology just blows me away! There was even some dialog during the music on the show - didn't matter.

How many 10s of thousands of hours (if not 100s of thousands) must there be of recorded music, including oodles of classical music? I just don't understand how the searches can be so fast.

Shazam made me feel real provincial - was watching a terrible Thai rock-em sock-em movie and one of the female characters started singing. Drives my gal nuts that I listen with the sound up to dialog I don't understand, so of course I had to rave about the incomprehensible tune and how deeply the words moved me. Insisted she Shazam the tune - she did, and had a pretty instant result, and was able to find an album cover that had the singing character's name and face.

Had to recognize that Shazam wasn't just capable of identifying English language tunes, but ALL recorded music - worldwide. Every now and then I feel real small.

 

[Options]

+1 to Shazam. And I'll take an android over an iphone any day of the week, personally.

 

[Ian S]

Not really new - I remember a friend demonstrating Shazam a number of years ago - but still quite amazing.

 

[NW-Bound]

There was a song that was special to me because a personal emotional experience back when I was 18. For 30 years, I looked for it to no avail. Only about 10 years ago, I happened to spot it on youtube as an upload by a Japanese girl!

I was able to research further, and found out that it was a 1973 one-hit wonder by a German singer. It enjoyed some popularity in Europe in that time, and had lyrics in many different languages, though I happened to listen to it in English.

Nowadays, I would not have to spend 30 years to track down a song.