CODEWITHSUNDEEP
pythonspeech-recognitionspeech-to-textpsychopyjulius-speech
WhooNo
WhooNo

Reputation: 990

adintool or psychopy for capturing speech for recognition by Julius?

I am working with Julius to recognize speech. I would like to have speech passed to Julius consistently as long as the mic is running, and store individual words to a library. I am looking for the best method to record and/or pass audio only when there is speech detected with the lowest use of memory and data.

adintool is developed for Julius. Would code like this:

(sender) % adintool -in mic -out adinnet -server receiver_hostname
(receiver) % julius -C ... -input adinnet

suggest that using adintool would allow me to pass the audio to Julius in server mode?

PsychoPy is Python based, and while I'm more familiar with Python Julius is written in C. PsychoPy would record sound and send it to Julius as a WAV file. This makes me suspect that it will lean heavier on memory and data resources.

This thread has some useful suggestions for recording only when there is a sound. Would it make sense to use this before either adintool or PsychoPy? Or do those tools already have this capability?

SUMMARY: Between adintool and PsychoPy, which is the best for lightweight audio capture for processing by julius, and would additional code be required to record/capture only when there is sound?

Upvotes: 2

Views: 159

Answers (1)

Nikolay Shmyrev
Nikolay Shmyrev

Reputation: 25220

You'd better use something like https://pythonhosted.org/pyjulius/ from Python. Julius server will record audio by itself and just send you back the transcribed text.

Upvotes: 1

Related Questions