In theory Spinx is exactly what I want. I want speech recognition with a limited result set so that I can maximize my accuracy.
As an example I may want someone to say a number and then a key word to say done.
Like 1234 Done.
I will then process the 1234 as my result and move on. So Done would be my KEYPHRASE to say stop listening.
In using the test app though, I managed to change my KEYPHRASE, but if I just sit here in a noisy office, it keeps hearing my phrase "BEAM ME UP". Am I doing something wrong? Why does it keep returning that, even when there is no talk, just noise? Nuance had a confidence value, is there something similar with Sphinx?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
In theory Spinx is exactly what I want. I want speech recognition with a limited result set so that I can maximize my accuracy.
As an example I may want someone to say a number and then a key word to say done.
Like 1234 Done.
I will then process the 1234 as my result and move on. So Done would be my KEYPHRASE to say stop listening.
In using the test app though, I managed to change my KEYPHRASE, but if I just sit here in a noisy office, it keeps hearing my phrase "BEAM ME UP". Am I doing something wrong? Why does it keep returning that, even when there is no talk, just noise? Nuance had a confidence value, is there something similar with Sphinx?
Yes, there is kwsThreshold in sources, you can see it if you read the Android demo code.