After filling out the form on the Dolby site, you can download the code right away. Nobody sits there approving any organization or person.
Checking the "Dolby Dialogue Intelligence" code, it's very well documented pure ANSI C.
It converts the input to 16kHz, runs it through a bunch of detectors and combines the results to get a value on whether speech is there or not.
The algorithm produces a 0 or 1 per sample, so that's a per-sample gate applied to the audio stream. That response has a set delay too, which makes it easier to compensate. The details are more difficult to fathom for an amateur coder like myself.
The gated audio is then fed to the 1770.1 measuring algorithm.
Basic structure seems simple. I'm just not sure we can share the source code like is done with the whole SWS extension.
Last edited by airon; 09-22-2018 at 01:18 PM.
|