Listen to sound

[Medic873]

Is there any way for Auto it to reconize part of a image like the first 20 x 20 pixels in a image it can reconize and then 20-30 x 20-30 it will reconize sepreatly or could i have it reconize a sound the sound would be easiest so it would reconize a certin sound if you could point me in the correct direction for this that would be great.

[jchd]

The search tool should be able to turn up example code or ready to use functions for extracting a fixed part of an image. I can't give you pointers as I'm surely no graphics guru.

Then I'd compute once for all the hash (MD5 is very fast and will be perfect) of the reference image. Then for each "candidate" image, extract the same part, compute its hash and compare hashes.

I use this to "recognize" thumbnails from a website which displays text in graphics form in a futile attempt to escape automated processing. I keep both the thumbnails and their MD5 in a database. It's very fast and reliable. For those who ask: no that isn't a gaming site.

The same approach will work for sound as well, provided you can intercept or get read access to the binary contents of the sound.