search

respeaker / respeaker_python_library

To build voice enabled objects/applications with Python and ReSpeaker
Apache License 2.0
149 stars 74 forks source link

Wake word is not working #36

Open titusece opened 6 years ago

titusece commented 6 years ago

Hello, I have edited the files like below to add the wake word, seems alexa itself is not working for me and new 'hello' word. Able to detect the 'respeaker' only. Can you please help with that ?

pi@raspberrypi:~/respeaker_python_library/examples $ cat ../respeaker/pocketsphinx-data/dictionary.txt 
respeaker   R IY S P IY K ER
alexa   AH L EH K S AH
play    P L EY
music   M Y UW Z IH K
hello   HH AH L OW
pi@raspberrypi:~/respeaker_python_library/examples $ cat ../respeaker/pocketsphinx-data/keywords.txt 
respeaker /1e-30/
alexa /1e-30/
play music /1e-40/
hello   /1e-30/
pi@raspberrypi:~/respeaker_python_library/examples $
xiongyihui commented 6 years ago

Maybe have a try to decrease the threshold:

hello /1e-40/
titusece commented 6 years ago

Thank you. Let me try and update you.

premanandmanimaran commented 5 years ago

Wake word is not working even after decreasing the threshold. I have updated the Keyword file too. When i try it with respeaker wake word it works, but when i change the wake word to custom word its not working. Do i need to update this anywhere else? Also How to make the microphone more sensitive ? Cause i have to go very close even with default 'respeaker' wake word.

Please help Xiongyihui

Prem

theronic commented 5 years ago

What to do after changing wake words, do you need to restart some service?

xiongyihui commented 5 years ago

@theronic If the recognition program is started, a restart is required.

@premanandmanimaran If your keyword can not be detected, you can adapt your voice with current acoustic module. See http://cmusphinx.sourceforge.net/wiki/tutorialadapt. It will effectively improve individual keyword recognition.