Здравствуйте Николай,
у меня вопрос по поводу того, как можно ускорить процесс распознавания - а
именно чтобы уменьшить ожидание паузы в конце речи с одной секунды до
чегонибудь поменьше. Мне нужно сделать так, чтобы не обязательно дожидаться
конца всей речи - а просто прекращать запись речи даже если небольшая пауза
(между словами) найдена. Я пробовал в pocketsphinx_continuous.c менять
значение с секунды до 0.1:
if ((cont->read_ts - ts) > DEFAULT_SAMPLES_PER_SEC*0.1)
break;
но все осталось как прежде. Может я не правильно сделал?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Nickolay,
i want some help about the utterances - i need to make lower the waiting for
the utterances - from 1 second to about 0.1. I need, in other words, not
necessary to wait for the end of the whole phrase - at least to wait for small
utterance between two words for example. I tried to change in
pocketsphinx_continuous.c the value to 0.1:
if ((cont->read_ts - ts) > DEFAULT_SAMPLES_PER_SEC*0.1)
break;
but the result seem to be the same, did i do smth wrong?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I suggest you to have more formal requirement first. It's not quite clear what
do you mean by "waiting". Once you make it measurable you can optimize it.
VAD has internal time range where it can allow switch from speech state to
silence state. This time range should be also accounted.
If you need immediate response, you need to plug into decoding process, and
you should not wait for the utterance end. For example you can track partial
results.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Здравствуйте Николай,
у меня вопрос по поводу того, как можно ускорить процесс распознавания - а
именно чтобы уменьшить ожидание паузы в конце речи с одной секунды до
чегонибудь поменьше. Мне нужно сделать так, чтобы не обязательно дожидаться
конца всей речи - а просто прекращать запись речи даже если небольшая пауза
(между словами) найдена. Я пробовал в pocketsphinx_continuous.c менять
значение с секунды до 0.1:
но все осталось как прежде. Может я не правильно сделал?
Hi Nickolay,
i want some help about the utterances - i need to make lower the waiting for
the utterances - from 1 second to about 0.1. I need, in other words, not
necessary to wait for the end of the whole phrase - at least to wait for small
utterance between two words for example. I tried to change in
pocketsphinx_continuous.c the value to 0.1:
but the result seem to be the same, did i do smth wrong?
I suggest you to have more formal requirement first. It's not quite clear what
do you mean by "waiting". Once you make it measurable you can optimize it.
VAD has internal time range where it can allow switch from speech state to
silence state. This time range should be also accounted.
If you need immediate response, you need to plug into decoding process, and
you should not wait for the utterance end. For example you can track partial
results.
Hi Nickolay,
you seem to show me that very thing i need - but how can i track partial
results? what i need to do?
Partial result can be retrieved with the same ps_get_hyp as final one. You
don't need to call ps_end_utt before that.