From Newsgroup: alt.os.linux.ubuntu
On 3/19/2024 10:05 AM, pinnerite wrote:
Julius speech recognition
Has anyone tried it?
Can you point me to the setup/usage instructions?
TIA Alan
You'll remember that HAL would not open the pod bay doors.
"Sudo make me a sandwich"
That doesn't work either.
It helps a lot to be in a quiet room (anechoic chamber)
with a good microphone. This is the single most popular
failure mechanism for speech recognition to not work.
I have a good microphone now, but... not the quiet room
I would need.
It's possible to isolate sound with an array microphone.
These MUST come with good software, as a pile of hardware
by itself, is not going to make anyone a sandwich.
https://www.minidsp.com/products/usb-audio-interface/uma-16-microphone-array
A lot of experiments we do on computers, are "GIGO".
Garbage In, Garbage Out. That's why the quality of the
input matters so much. I've attempted speech recognition
before, where the "test interface" told me there was
"no audio signal present". That's important, by the way :-)
Who knew.
I think now you can tell why I have not, and will not
be testing this. I've learned enough over the years
about "GIGO" to have given up. This is why devices
like Alexa use an array setup. The commercial companies
have figured out the price of admission.
Paul
--- Synchronet 3.20a-Linux NewsLink 1.114