First of all I would like to say that I do not have much experience in this field but am more than willing to learn. I am working on a project that uses pocketsphinx_continuous to recognize a few commands (I have narrowed down the list with a grammar file) on windows. I made it work properly but I have been changing the original code to better fit my needs, the thing is (if I understand it correctly) pocketsphinx_continuous reads from the microphone until there is more than one second of silence. This inevitably slows down the processing of the word for one second which could be crucial so I have been thinking of implementing some Push-To-Talk action. Not starting the loop without the button being pressed is easy, but how could abruptly end the process? If I decrease the one-second tolerance it might be so that the loop ends before the person starts to speak.
I initially thought of checking the state of the button before each iteration of the loop and then "break" whenever it was released but it failed sometimes because it was not able to find the end of the utterance. Could anyone suggest another solution? If I were to keep this one, would it be possible to change the software so that if it doesn't find any end nodes it would use the last one as end node?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
You can use keyword spotting mode and check for hypothesis without waiting for silence. The response time will be less than 0.1s then and you will also be able to ignore other words.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello everyone,
First of all I would like to say that I do not have much experience in this field but am more than willing to learn. I am working on a project that uses pocketsphinx_continuous to recognize a few commands (I have narrowed down the list with a grammar file) on windows. I made it work properly but I have been changing the original code to better fit my needs, the thing is (if I understand it correctly) pocketsphinx_continuous reads from the microphone until there is more than one second of silence. This inevitably slows down the processing of the word for one second which could be crucial so I have been thinking of implementing some Push-To-Talk action. Not starting the loop without the button being pressed is easy, but how could abruptly end the process? If I decrease the one-second tolerance it might be so that the loop ends before the person starts to speak.
I initially thought of checking the state of the button before each iteration of the loop and then "break" whenever it was released but it failed sometimes because it was not able to find the end of the utterance. Could anyone suggest another solution? If I were to keep this one, would it be possible to change the software so that if it doesn't find any end nodes it would use the last one as end node?
You can use keyword spotting mode and check for hypothesis without waiting for silence. The response time will be less than 0.1s then and you will also be able to ignore other words.