I'm begginner in CMUSphinx4.
I need to comparate two samples of voice recognition.
The first is a sample of people native to American English (The pattern).
The second is a sample of people from another language.
How to evaluate the recognition of both?
Thanks in advance!!!
This problem is calleed pronunciation evalution, and currently there is no existing solution either in sphinx4 or in any other open source speech recognition framework.
You can read more about that problem and possible ways to tackle it in CMUSphinx blog. The most difficulty stems from the nature of human language understanding. There is no a single reference speaker to compare with, and ultimately all the people have variation in their pronunciation.