A bit random question (but I’m not trolling, I’m just an artist); is there a way to train Mycroft to wake on a sound that is not a word? Imagine I trained on a dataset of mewing cats?
I understand Mimic is a speech recognition engine but relatively comfortable with Myroft’s backend by now I’m trying to push further.
yes that should work. there is also a pricise model with a cough. you probably think about it when the cat screams the door opens or there is food. look at how to train precise. https://github.com/MycroftAI/mycroft-precise/wiki/Training-your-own-wake-word#how-to-train-your-own-wake-word
not stupid at all, mycroft did it with coughing sounds
https://www.bizjournals.com/kansascity/news/2018/10/17/sickweather-mycroft-coughing-sensors.html
Note that mimic is used to generate speech not recognize , what you are looking for is precise
https://mycroft-ai.gitbook.io/docs/mycroft-technologies/precise
Indeed, it can hypothetically be done with any number of sounds. Google, a few months ago, created wake words for the various sounds of potential catastrophes. Windows breaking, smoke alarms, that sort of thing, the idea being it would text you if your home caught fire.
Then they deployed it - ostensibly by accident - without telling anyone! Big scandal.
Anyway, if they can do it, so can other listeners. It’s just another waveform. The computer can’t tell what the audio represents, just that it matches the desired audio.
Thanks everyone for the tips; indeed now makes sense. I will try to train a new one and post how it goes! We might also try to do some funky ones with some of my students, will share in case something nice comes out! Hope you have great weekends!