Computing.Net > Forums > iPhone > How does 'Shazam' work?

Computer Problems? Computing.Net has over 1,000,000 posts about all things technology related! Over 90% answered within 24 hours! Click here to start participating now! Also, be sure to check out the New User Guide.

How does 'Shazam' work?

Reply to Message Icon

Name: justinblue
Date: December 30, 2008 at 05:54:38 Pacific
OS: iPhone OS
Subcategory: iPhone
Comment:

This enquiring mind wants to know!

It seems like magic.

But something tells me it isn't.

--

I love words.
- Susannah McCorkle

archived from misc.phone.mobile.iphone



Sponsored Link
Ads by Google

Response Number 1
Name: justinblue
Date: December 30, 2008 at 06:54:38 Pacific
Reply:

On Dec 30, 5:54=A0am, MC <copes...@mapca.inter.net> wrote:
> This enquiring mind wants to know!
>
> It seems like magic.
>
> But something tells me it isn't.
>
> --
>
> I love words.
> =A0- Susannah McCorkle

Shazam analyzes the song and uses an acoustic fingerprint that it
compares to a music database to determine a match.

It could generate an acoustic fingerprint using any of the following
well known audio features: Fourier coefficients, Mel Frequency
Cepstral Coefficients (MFFC), spectral flatness, sharpness, Linear
Predictive Coding (LPC) coefficients and others. Also derived
quantities such as derivatives, means and variances of audio features
could be used. The extracted features are mapped into a more compact
representation by using classification algorithms, such as Hidden
Markov Models, or quantization=85

Does that answer your question Susannah?

archived from misc.phone.mobile.iphone


0

Response Number 2
Name: justinblue
Date: December 30, 2008 at 07:54:38 Pacific
Reply:

In article
<b816ca1f-816a-4180-b28c-efc72e3d79d3@z6g2000pre.googlegroups.com>,
4phun <vic.healey@gmail.com> wrote:

> Shazam analyzes the song and uses an acoustic fingerprint that it
> compares to a music database to determine a match.
>
> It could generate an acoustic fingerprint using any of the following
> well known audio features: Fourier coefficients, Mel Frequency
> Cepstral Coefficients (MFFC), spectral flatness, sharpness, Linear
> Predictive Coding (LPC) coefficients and others. Also derived
> quantities such as derivatives, means and variances of audio features
> could be used. The extracted features are mapped into a more compact
> representation by using classification algorithms, such as Hidden
> Markov Models, or quantizationŠ
>
> Does that answer your question Susannah?

It does... but I'm not Susannah.

Thanks!

--

I love words.
- Susannah McCorkle

archived from misc.phone.mobile.iphone


0

Response Number 3
Name: justinblue
Date: December 30, 2008 at 08:54:38 Pacific
Reply:

On 2008-12-30, 4phun <vic.healey@gmail.com> wrote:
> Does that answer your question Susannah?

It doesn't answer the question of how on earth they get a copy of
every-song-ever-released in order to get a fingerprint of them in
the first place ;-)

archived from misc.phone.mobile.iphone


0

Response Number 4
Name: justinblue
Date: December 30, 2008 at 09:54:38 Pacific
Reply:

On Dec 30, 10:34=A0am, Jon Ribbens <jon+use...@unequivocal.co.uk> wrote:
> On 2008-12-30, 4phun <vic.hea...@gmail.com> wrote:
>
> > Does that answer your question Susannah?
>
> It doesn't answer the question of how on earth they get a copy of
> every-song-ever-released in order to get a fingerprint of them in
> the first place ;-)

pirate bay - torrents?

archived from misc.phone.mobile.iphone


0

Response Number 5
Name: justinblue
Date: December 30, 2008 at 10:54:38 Pacific
Reply:

On Tue, 30 Dec 2008 09:34:27 -0600, Jon Ribbens
<jon+usenet@unequivocal.co.uk> wrote:

>On 2008-12-30, 4phun <vic.healey@gmail.com> wrote:
>> Does that answer your question Susannah?
>
>It doesn't answer the question of how on earth they get a copy of
>every-song-ever-released in order to get a fingerprint of them in
>the first place ;-)

I would guess that people who have any interest in selling their song
would be glad to make fingerprints available.

archived from misc.phone.mobile.iphone


0

Related Posts

See More



Response Number 6
Name: justinblue
Date: December 30, 2008 at 11:54:38 Pacific
Reply:


"4phun" <vic.healey@gmail.com> wrote in message
news:b816ca1f-816a-4180-b28c-efc72e3d79d3@z6g2000pre.googlegroups.com...
On Dec 30, 5:54 am, MC <copes...@mapca.inter.net> wrote:
> > This enquiring mind wants to know!
> >
> > It seems like magic.
> >
> > But something tells me it isn't.
>
>
>
> Shazam analyzes the song and uses an acoustic fingerprint that it
> compares to a music database to determine a match.
>
> It could generate an acoustic fingerprint using any of the following
> well known audio features: Fourier coefficients, Mel Frequency
> Cepstral Coefficients (MFFC), spectral flatness, sharpness, Linear
> Predictive Coding (LPC) coefficients and others. Also derived
> quantities such as derivatives, means and variances of audio features
> could be used. The extracted features are mapped into a more compact
> representation by using classification algorithms, such as Hidden
> Markov Models, or quantizationS

So, essentially, per Arthur C. Clarke's definition, it's magic. ;-)

archived from misc.phone.mobile.iphone


0

Response Number 7
Name: justinblue
Date: December 30, 2008 at 12:54:38 Pacific
Reply:

In article
<b816ca1f-816a-4180-b28c-efc72e3d79d3@z6g2000pre.googlegroups.com>,
4phun <vic.healey@gmail.com> wrote:

> On Dec 30, 5:54 am, MC <copes...@mapca.inter.net> wrote:
> > This enquiring mind wants to know!
> >
> > It seems like magic.
> >
> > But something tells me it isn't.
> >
> > --
> >
> > I love words.
> >  - Susannah McCorkle
>
> Shazam analyzes the song and uses an acoustic fingerprint that it
> compares to a music database to determine a match.
>
> It could generate an acoustic fingerprint using any of the following
> well known audio features: Fourier coefficients, Mel Frequency
> Cepstral Coefficients (MFFC), spectral flatness, sharpness, Linear
> Predictive Coding (LPC) coefficients and others. Also derived
> quantities such as derivatives, means and variances of audio features
> could be used. The extracted features are mapped into a more compact
> representation by using classification algorithms, such as Hidden
> Markov Models, or quantization…
>
> Does that answer your question Susannah?

What's most interesting about it is that it doesn't require the song to
be played at the beginning, so it must have the capability of searching
through songs at least to some degree. Most "songs" have a AABA
pattern, where the beginning part is different from the "bridge" part,
sometimes both in key an tempo.

There has to be someway of syncronising the music "snippet" that's
sampled off the iphone with the music library. Conceivably, one could
analyze the music looking for downbeats and sync on those, but that
might be hard with some music that doesn't have an easily recognizable
downbeat (like say, pizzicato sections).

As far as I know, it only finds things that are in the iTunes music
library, so they have a vast supply of songs to refer to. It also
seems to work better with vocals - having more trouble on the
instrumental only portion. Maybe there's something in their Spotlight
techology that they can use to get digital fingerprints of segments of
the music.

-Owen

archived from misc.phone.mobile.iphone


0

Response Number 8
Name: justinblue
Date: December 30, 2008 at 13:54:38 Pacific
Reply:

On 2008-12-31, O <owenx@denofinequityx.com> wrote:
> As far as I know, it only finds things that are in the iTunes music
> library

I think that's extremely unlikely - Shazam has been around since 2002,
whereas the iTunes Music Store has only been around since 2003. There
is no link I am aware of between Apple and Shazam, the iPhone app is
just one of several ways to access their service.

archived from misc.phone.mobile.iphone


0

Response Number 9
Name: justinblue
Date: December 30, 2008 at 14:54:38 Pacific
Reply:

In article <slrnglnsq3.te1.jon+usenet@snowy.squish.net>, Jon Ribbens
<jon+usenet@unequivocal.co.uk> wrote:

> On 2008-12-31, O <owenx@denofinequityx.com> wrote:
> > As far as I know, it only finds things that are in the iTunes music
> > library
>
> I think that's extremely unlikely - Shazam has been around since 2002,
> whereas the iTunes Music Store has only been around since 2003. There
> is no link I am aware of between Apple and Shazam, the iPhone app is
> just one of several ways to access their service.

The only reason I said that is that there's a link on the iPhone Shazam
to get the song from the iTunes Music Store. Having looked at their
web page, you're probably correct in that there's probably not a
connection.

-Owen

archived from misc.phone.mobile.iphone


0

Response Number 10
Name: justinblue
Date: December 30, 2008 at 15:54:38 Pacific
Reply:


"O" <owenx@denofinequityx.com> wrote in message
news:311220082104141727%owenx@denofinequityx.com...
> In article <slrnglnsq3.te1.jon+usenet@snowy.squish.net>, Jon Ribbens
> <jon+usenet@unequivocal.co.uk> wrote:
>
>> On 2008-12-31, O <owenx@denofinequityx.com> wrote:
>> > As far as I know, it only finds things that are in the iTunes music
>> > library
>>
>> I think that's extremely unlikely - Shazam has been around since 2002,
>> whereas the iTunes Music Store has only been around since 2003. There
>> is no link I am aware of between Apple and Shazam, the iPhone app is
>> just one of several ways to access their service.
>
> The only reason I said that is that there's a link on the iPhone Shazam
> to get the song from the iTunes Music Store. Having looked at their
> web page, you're probably correct in that there's probably not a
> connection.

I use the Cingular version of Shazam on my Windows Mobile phone, and it
links to AT&T's Napster WAP page to purchase instead. I suspect each
version of Shazam simply links to whoever's service that version of Shazam
is customized for, with Shazam getting whatever appropriate sales commission
is offered.

archived from misc.phone.mobile.iphone


0

Response Number 11
Name: justinblue
Date: December 30, 2008 at 16:54:38 Pacific
Reply:

Jon Ribbens <jon+usenet@unequivocal.co.uk> wrote:

> how on earth they get a copy of
> every-song-ever-released in order to get a fingerprint of them

they don't.
I've got dozens of songs which are not recognized by Shazam (and also
not by Midomi). Most of those are without human voices...

Hmm, I have yet to find an instrumental song which IS recognized at all.

--
In a world without walls and fences,
who needs windows and gates?

archived from misc.phone.mobile.iphone


0

Sponsored Link
Ads by Google
Reply to Message Icon






Post Locked

This post is quite old and has been locked from receiving new replies. Please create a new posting instead.


Go to iPhone Forum Home


Sponsored links

Ads by Google


Results for: How does 'Shazam' work?

MP3s www.computing.net/answers/phones/mp3s/709.html

New Palm Pre won't work with old ap www.computing.net/answers/phones/new-palm-pre-wont-work-with-old-ap/285.html

Does Push work for you? www.computing.net/answers/phones/does-push-work-for-you/774.html