[asterisk-users] Identify remote prompts: Partial audio matching?

[Philipp von Klitzing]

> You might be able to record these snippets then pass them through the
> Vestec or Lumenvox Speech engine to get what you want. 

Unfortunately that won't work because:

* the containing recordings/feeds can be quite long, can be 
embedded/surrounded by silence, ringing tones, music or special tones, 
and the ASR engines are not really designed to handle this situation. 

* next to this both LumenVox and Vestec do not cover the language(s) that 
I need this for, since both companies are focused on the American market 
(and yes, I am aware of Loquendo and Nuance).

So that is why I am looking for something like partial audio 
fingerprinting; this is a bit like these "find duplicate mp3 songs in my 
huge media library" tools, only that in this case it is 1. not about an 
exact duplicate, and 2. the audio quality can vary, and 3. this is about 
finding contained parts instead of comparing full songs with each other.

Philipp