ENMFP and echoprint-server

106 views
Skip to first unread message

Mirosław Sztorc

unread,
Mar 26, 2014, 9:28:48 AM3/26/14
to echo...@googlegroups.com

Is possible to store and search ENMFP hashes (codes) in echoprint-server?

David DesRoches

unread,
Mar 26, 2014, 10:35:38 AM3/26/14
to echo...@googlegroups.com
Miroslaw,

No. The structure of ENMFP hash codes is different from Echoprint. An Echoprint string includes a concatenated series of 8 sub-bands and is collated into 60-second segments for indexing in SOLR.

David


On Wed, Mar 26, 2014 at 9:28 AM, Mirosław Sztorc <mirek....@gmail.com> wrote:

Is possible to store and search ENMFP hashes (codes) in echoprint-server?

--
You received this message because you are subscribed to the Google Groups "echoprint" group.
To unsubscribe from this group and stop receiving emails from it, send an email to echoprint+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.



--
David DesRoches
Lead Audio Engineer
The Echo Nest

Mirosław Sztorc

unread,
Mar 26, 2014, 11:22:36 AM3/26/14
to echo...@googlegroups.com
Thank you David - I understand.

I'm interested ENMFP because I want to reduce the record time required to detect audio. How can I modify algorithm of echoprint-codegen to increase precision of detecting short pieces of audio assuming that every record is a good quality (without noise and distorsion)? I know this algorithm is dedicated to average time of song and  over-the-air needs, but duration of my files in database is max 60 sec and I have to search using short pieces (5-6 secs).

What's is your opinion about this problem?

David DesRoches

unread,
Mar 27, 2014, 10:54:27 AM3/27/14
to echo...@googlegroups.com
Miroslaw,

Others in the group may have done experiments with this, but my sense is that you would need considerably more density in the fingerprint codes to get a credible match with only five seconds of audio. This places much greater storage and CPU burden on the server. Since the Echoprint vision was to match songs no matter which portion was fingerprinted on the client, the server has to index entire tracks, so increased density drives up the server resources rapidly.

David


Mirosław Sztorc

unread,
Mar 29, 2014, 5:36:13 AM3/29/14
to echo...@googlegroups.com
David,

Thank you for your response. Assuming that resources do not matter, how can I increase a density of fingerprinting using codegen?

Miroslaw
Reply all
Reply to author
Forward
0 new messages