How to fingerprint a television episode

[Richard Penner]

I want to build a platform which uses an iOS app to show a "second screen" for a particular television show. The app involves displaying different information at different times throughout the show. The app experience must be passive, in that I'd like the user to just hit "start", and it syncs up automatically, no pausing needed if the video is paused. 

Perhaps I'm crazy but it looks like an Echonest server might work. What I'm specifically curious about is how to fingerprint a television episode. My goal is to be able to launch the app at any given point in any episode, and it figures out which episode you are watching (nice to have), and precisely where in that episode you are, with about an accuracy of one second or so (required).

I'm new to audio fingerprinting, so

1) should my echonest server fingerprint the entire episode as one "song"? Any reason to split it up?

2) when my echonest client does a search and finds a match, am I told "where" it matched (eg how far into the "song" / episode)? If not, is there some other means to figure this out?

Any help would be appreciated! Thank you,

Richard

[S Mallikarjun]

Hi Richard,

you may not get the exact point "where" it matched but you can figure out close(in 60 seconds scale)

each song/episode in echonest will be split in to 60 sec duration which is called doc and multiple docs are ingested in to database .each song/episode will have multiple docs.

unfortunately with existing api you can retreive doc id.

You can modify the server if you plan to host your server .

Modify best_match_for_query in fp.py to get doc id

Regards
Mallik

Regards
Mallik

--
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/groups/opt_out.
 
 

[Richard Penner]

Thanks Mallik.

If I split the audio into 60 second docs and match against them, do I have to worry about the fingerprint spanning across a doc boundary? 60 seconds unfortunately won't be accurate enough.

I found in this group another thread where someone had matched a 5 second sample against 15 second tracks. Does this scale if I split the whole episode into 15 second docs, or would the matching accuracy decrease with that many short docs?

I will be hosting my server, so thanks for the tip about modifying the server.

Richard

[S Mallikarjun]

Did nt try with 15sec yet.but i think fingerprint match takes more time.