Episode 464: Particle Voice Recognition for Home Appliances

https://www.element14.com/community/people/aspork42/blog/2020/10/12/building-a-smart-home-audio-notification-system

I just posted this project using a library for the ESP8266 which can do text to speech via an MQTT message. The pronunciation and quality is pretty poor but it works. I ended up using an online speech synthesis system and saving pre-recorded mp3s for the announcements that don't require "on the fly" text to speech.

The video in the post only shows the mp3; not the text to speech

Anyone have any Good Text to Speech products/libraries/examples?

That's cool! I have found out the hard way that despite my iPhone having great background noise cancelation, that must be more of a software feature. When calling from apps like Teams or WhatsApp, the mic works poorly and amplifies background noise.

So the module you have must be a little /too/ simple and hearkens to the early days of voice assistants where one had to repeat themselves 5 times and say things /juuuuuust/ right for it to work.

Good idea, I will try that tomorrow while I try the other suggestions from Mike and Dave.

-Kaleb

Nice idea to simplify life.  Maybe placing the microphone/photon further away, maybe outside the laundry romm would make the background noise less of a problem? 

I'm not sure why the voice recognition module doesn't have a link, so hey, I know how to use duckduckgo, so I'll post it.

Speak Recognition, Voice Recognition Module V3 [AD_VR3] - $28.00 : Elechouse, Arduino Play House

The price is reasonable.

kmikemoo, @beacon_dave

Good ideas, I did not think of that. The last eight notes are a unique pattern (yes, I still have that tune stuck in my head!), I can retrain on those last eight and see if it works.

Ill let you guys know how that goes.

Thanks again

-Kaleb

Thanks DAB, It was an interesting and fun try.

-Kaleb

kalebclark  Brilliant application.  You mention that the dryer song repeats itself.  If the logic is "turn the dryer OFF", "play the song", I'm with beacon_dave on sampling later in the song.  I'm guessing that the logic must be "play the song" x times, "turn the dryer OFF".  Crazy thought... simple high-pass filter on your microphone input.  Maybe that will simplify the sample to the point where the background rumble doesn't affect it.

Did you try sampling on the end part of the tune rather than the beginning ? It sounds like the dryer stops spinning down before the tune finally runs out, so may be able to match on the tune as it ends ?

Interesting project and module.

DAB