Implement on media monitoring (like radio ads monitoring)

Question

I want to develop an Audio Monitoring Software, for example to know how many ads of certain company where published on x radio station ?

There is any way to analyse "realtime" the audio stream and detect when any version of an ad is played on the radio?

Or the best way is to analyse every x seconds the audio fragment, if this is the way to go, what can I do to know if only a segment of an audio has the sample audio (for example analyse 20 minutes of radio and return true if the spot (ad) where player in that audio sample)

(Sorry for my English, I hope is understandable)

Answer 1

I guess realtime could be difficult due to the fact that you have to analyze your radiostream. For that you need to cache, analyze / fingerprint and run against an existing database.

But take a look on these questions:

https://stackoverflow.com/questions/2462410/acoustic-fingerprint-opensource

Musicbrainz fingerprinting

More Links:

http://acoustid.org

https://musicbrainz.org/doc/Fingerprinting

http://echoprint.me // service by spotify / echonest

https://www.audiblemagic.com/broadcast-infrastructure

Good luck.

Answer 2

An excellent open-source audio fingerprinting library in Python can be found here:

http://github.com/worldveil/dejavu

It allows you to fingerprint an audio file once, store the fingerprints in a database, and do continual recognition and adding fingerprints as time goes on.

You can even fingerprint small sections of song to save disk storage if you are just doing on-disk deduplication.