Background on a use case that will ultimately not work with Alexa as I had hoped. Spawning a separate thread to see if a existing or better path might exist.
I use a bedside alexa to set my morning alarm which isn’t constantly the same time. And I wanted to automate 30 minutes before the set alarm and not when the alarm goes off. Since I verbally set the alarm before I go to bed and alexa won’t expose the alarm set time, I have an idea which with the community, may already be solved. Thought I would run it by.
SO my idea is to use the cloud speech to text engines in parallel (or even in place) of Alexa. Something along these lines:
- Small linux device (pi, etc) with a microphone input sitting near where you give your alarm command.
- Write a program that listens over the mic 24/7, during a time period, based on a well trained key word, or by a button. Cost+Paranoia drives this decision.
- Send captured audio to one of several cloud API speech to text services.
- Receive text file response back which includes the time setting.
- Using a network service, pump your event over to smartthings device. App\device as needed to wrap the capability such that the event can be responded to.
There are many of cloud API services for this. Google is probably the popular choice, but there may be better\cheaper or ones that are less likely to hold on to your data. With good on\off control (training, button, otherwise), google is free up to 60 minutes and really cheap after that.
There are a bunch of python examples out there with code examples…some seem better with regard to normalizing noise and reducing audio submissions. This doesn’t eliminate the Alexa value and may only be beneficial for this use case, but you could certainly do Speech to text “smart control” for anything.
Thoughts? Simpler approach?