speech phonemes to letter (not dictation or speech-to-text)

edited July 2017 in Library Questions

My terminology may be lacking, please forgive me.

I would like to find a way to take audio in, recognize phonemes, and then use those phonemes within a processing sketch. The sketch does not need to understand the speech as text.

if you 'grrr' like a dog, I would like to have 'g-r-r-r-r' on the screen.

if you 'shhhh' it will 'sh-sh-sh-sh'

if you 'meow' it will 'm-ee-ow' or some such.

This is the beginning sequence of most speech recognition libraries, but I am finding that most no longer allow access to the 'under-the-hood' mechanics anymore.

The closest I came to this, while trying to find the similar thing for openframeworks, was Microsoft speech recognition, which was tied to the Kinect library, but of course, required windows and visual studio to work. Lacking a windows computer, I could not use this (and the library only supports the newest Kinect For Windows hardware now) This library DID allow access to the 'best guess' which I 'think' involved the phonemes.

I am more interested in a visual representation of the sound, but not necessarily the specificity of words and sentences.

Is there a library available for processing that does this? Someone mentioned an artwork that Golan Levin may have done which was similar to this (a poet speaking and their phonemes projected on a wall) but I have not found anything in his archive of works...

Tagged:

Answers

Sign In or Register to comment.