I have a question about dataset in mycroft-precise. I want to know what audio should be in the dataset .What kind of audio should be included in not-wake-word folder and wake-word folder?
Anything that’s not your wake word.
For mine I have included the google commands dataset, the open source noises dataset, over 500 clips of “noises” (air conditioner starting, coughs, sneezes), and 4k+ not-wake-words. The majority of my not-wake-words are from rhymes and similar sounding stuff, some are from words with similar phonemes.
This page needs a bit of updating about the training steps, but the dataset bits are still relevant.
Thank you, sir.
Do you have wake-up words with noise in your wake-word folder?
I recorded my wake words a number of ways, a few had noises in the background.
Thank you a lot. Do you have some contact information which I can communicate and learn with you.
You’re using it already. This or the chat room are the best places to do so.