phone_estimate.h File Reference

Detailed Description

declaration of the phoneme timing estimator to improve results from SAPI 5.1

SAPI can use various recognition engines, some from 3rd parties. If You use "Microsoft English Recognizer 5.1" then the default set, supplied in the code will work. Otherwise, you'll need to figure out the phoneme labels of your speech recognizer and create a new set of spacing rules here.

So that we can view the results in the lipsync tool, we also provide a way to map phonemes into the anno 40 set. I like being able to view the results.

Each phoneme mapper object has the engine phoneme, the output phoneme, an optional descripton, and a phoneme type.

Since SAPI 5.1 does not provide alignments for phonemes within a word, we have to do it ourself. The strategy is fairly simple, we spread out the phonemes according to the classification, where unvoiced phonemes have a short hold compared to voiced phonemes, dipthongs hold the longest.

Definition in file phone_estimate.h.

Go to the source code of this file.