Hello. Sorry for wrong title of the topic (I can't change it). My title was: "PocketSphinx recognizes words from dict during the speech incorrectly."
I have a dictionary and grammatics for pocketsphinx and it works. But when one person or several people talk near the microphone, pocketsphinx detects phrases from my grammatics. But people did not say any of this phrases.
In order to get rid of this problem, I have managed to use a key phrase in order to activate recognition of words from dictionary.
But use of key phrase does not help too: when one person or several people just talk near the microphone, the program somehow detects my key phrase. But it's wrong detection (in fact, people did not say my key phrase). It's a very big problem, actually.
Is there any additional settings in order to make pocketsphinx work better? Or it's impossible?
Last edit: Yurii 2016-04-15
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am trying to tune threshold using command
pocketsphinx_continuous -infile /myreccordedfile.wav -hmm /en-us/en-us -keyphrase "my pkeyphrase" -kws_threshold \"my threshpld" -time yes
but it gives error " my keyphrase" is missing in the dictionary .
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
You need to create your dictionary. Here is the tutorial. Example of my dictionary you can find here. But it's a Russian language.
Also you can find the official dictionary for English language here. But there are a lot of words, so you are able to chose words what you need in order to enhance the speed of recognition.
Last edit: Yurii 2016-04-19
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
NIckolay, can you please profide some information about theory of the keyword threshold? I think that it's better to know that it is, not just tune the value. Does it has a unit?
Last edit: Yurii 2016-04-30
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello. Sorry for wrong title of the topic (I can't change it). My title was: "PocketSphinx recognizes words from dict during the speech incorrectly."
I have a dictionary and grammatics for pocketsphinx and it works. But when one person or several people talk near the microphone, pocketsphinx detects phrases from my grammatics. But people did not say any of this phrases.
In order to get rid of this problem, I have managed to use a key phrase in order to activate recognition of words from dictionary.
But use of key phrase does not help too: when one person or several people just talk near the microphone, the program somehow detects my key phrase. But it's wrong detection (in fact, people did not say my key phrase). It's a very big problem, actually.
Is there any additional settings in order to make pocketsphinx work better? Or it's impossible?
Last edit: Yurii 2016-04-15
Each keyphrase has a threshold you need to tune to balance between detection and false alarms.
It's the only one way? If so, what is the max value of threshold?
Last edit: Yurii 2016-04-16
Thresholds could range from 1.0 to 1e-50.
Last edit: Nickolay V. Shmyrev 2016-04-16
I am trying to tune threshold using command
pocketsphinx_continuous -infile /myreccordedfile.wav -hmm /en-us/en-us -keyphrase "my pkeyphrase" -kws_threshold \"my threshpld" -time yes
but it gives error " my keyphrase" is missing in the dictionary .
You need to create your dictionary. Here is the tutorial. Example of my dictionary you can find here. But it's a Russian language.
Also you can find the official dictionary for English language here. But there are a lot of words, so you are able to chose words what you need in order to enhance the speed of recognition.
Last edit: Yurii 2016-04-19
but i recheked my dictionary and the keyphrase is available in my dictionary .then what to do in that case
NIckolay, can you please profide some information about theory of the keyword threshold? I think that it's better to know that it is, not just tune the value. Does it has a unit?
Last edit: Yurii 2016-04-30