Problems with commands recognition

[sesma]

I have configured the Rhasspy system with a few simple commands but it gives me a lot of problems when it comes to recognizing these commands, you have to repeat the command more than once to detect it correctly, what could it be? Is there any way to improve the system?

[GetTime]
what time is it
tell me the time

[GetTemperature]
whats the temperature
how (hot | cold) is it

[PercentageControl]
device_name = ((light | lights) {name})
place_name = ((kitchen | living room | all | house | home | room ) {name})
color_state = (green | red | blue | red | red) {color}

set <place_name> <device_name> [to] <color_state>
adjust <place_name> <device_name> [to] <color_state>
adjust <color_state> <place_name> <device_name>
set [to] <color_state> <place_name> <device_name>
adjust [to] <color_state> <place_name> <device_name>
tune <place_name> <device_name> [to] (0…100) [percent]
regulate <place_name> <device_name> [to] (0…100) [percent]
adjust to (0…100) [percent] <place_name> <device_name>
tune to (0…100) [percent] <place_name> <device_name>
regulate to (0…100) [percent] <place_name> <device_name>

[ChangeState]
light_name = ((living room light | kitchen light | room light) {name})
light_state = (on | off) {state}

turn <light_state> [the] <light_name>
turn [the] <light_name> <light_state>

[geoffrey]

Try to test what Rhasspy understands when you invoke it using text input and your voice.

On the Speech Tab you can first type the sentence and then select Get Intent to see if the result from the sentence is the expected intent with all entities tagged properly.

Afterwards wake Rhasspy up and say the exact same sentence. On the Speech Tab you can see if an intent was detected and also what the raw text was it understood.

[sesma]

Thank you very much, with this I have managed to fix some small bugs in my code. Even so, the command recognition as such is not very good, it does not detect all the commands perfectly. Is there any way to manually train the system, or any option to make it work better and detect the commands better?

[geoffrey]

You can also try to verify how the recorded voice command sounds like.

@synesthesiam described the command for recording earlier in another post:

Snowboy custom wakeword false positive

I usually just use arecord to record something on the Pi. Use arecord -L to a a list of available devices. A command like this will record it like Rhasspy does: arecord -D '<DEVICE>' -r 16000 -f S16_LE -c 1 -t wav > /path/to/my.wav Hit CTRL+C when you’re finished speaking.

Afterwards you can use aplay to play replay it.

Good luck.