The voice recognition project advance.
A library of several basic words is being created, the audio format will be. "Jo"
The technique requires an alsa api which gives us a sampled signal, the sampled signal is analyzed first to identify white How many forms of curve up the word, for example the word "hello" we find five "special sound curve" or "may be used for other words. Each curve is identified using my model mathematics using a dimensionless number (very light as it replaces thousands of points). One of the other forces of the library is that it will contain only possible with a foundation construction limited to certain words useful to the builder of the year, thus making it possible to eliminate all the noise and gain speed. The constructor / decoder file as an "jo" is planned at the moment in java (response to my tests), but the use I will fix the final language. The source code will be placed in one of my fields, that the manufacturer exercises.
Five minutes of laughter for an adult, had an eternity to a child.
0 comments:
Post a Comment