Only wakeup mechanism

Hi,

Is it possible to use only the wake up mechanism> can we disable the ASR usage and only use the wake up word detection?

Thanks,
Tomer

Hi @tomerpeled!

That’s totally possible, and you’re actually not the first person to ask about it recently.

To summarize that other post, you have a couple options:

  • Make a custom speech pipeline profile that omits the ASR stage (see one of our existing profiles for an example of how to do this) and pass it to your Spokestack.Builder via .withPipelineProfile("full.name.of.your.class").
  • Grab the pipelineBuilder from Spokestack.Builder and set the processing stages on it just like the profile classes do.

Either way, what you’re doing is configuring the speech pipeline to avoid sending audio through ASR. You’ll get the ACTIVATE / DEACTIVATE events from the wake word processor and whatever timeout mechanism you have in place, but you won’t get any RECOGNIZE events. I recommend including the ActivationTimeout stage and setting low wake-active-min and wake-active-max values to maximize sensitivity to the wake word.

Since this appears to be a useful feature, we’ll also consider adding a pre-made wake word-only profile to the library—contributions are welcome if you’re up for it!

Josh

I’ve created an issue for the new profile with a little more detail on the process.

Great! I’ll take a look on it.

Thanks!