TV audio processing with TV capture card

Question

Hello,

I'm looking for an open source library or framework to process audio signal from a TV capture card. The idea is to detect TV ad spots and register the time and the channel where them happends. I never worked in something like this, so, any information, link, idea is welcome.

Thanks in advance!

EDIT: I don't care about the language or the OS. I want to clarify that my objective is to detect the specifics ads of the company where I work. So, I think I should use the fingerprints of our ads and check them against the captured data. If you have information about this, is welcome too.

Answer 1

Its possible to use something like virtual dub to capture the video initially.

Detecting ads tends to be a combination of several things:

Ad lengths tend to be multiples of 5 seconds, so the suspected start and end points can be verified against this (can vary in different countries though).

Sound generally drops completely at the transition point.

A blank frame or a significant change occurs also at the same time.

Depending on station there could be other logos and things you can detect when ads start.

Obviously not all these things are guaranteed, and you tend to end up making a judgement based on all these factors, and have some kind of visual check at a later point.

Edit:

To find look for particular ads, you will need to store some pertinant information on the ad (as you say a fingerprint) such as:

Ad length.
Sounds levels, maybe taken at several points in time.
Color info, maybe rgb levels to start with, again as they vary throughout the ad.

As a combination of these things you can come up with 'candidates' for matches though by no means with 100% accuracy.