Non-specificity of voice recognition for voice assistance devices or features

I remember from quite some time ago when Siri was first launched that there was no specific way of getting the voice-triggered assistant to recognize specific voices.

These days there are many different home assistant devices like Amazon Alex and Google Home, and some information is provided in terms of setting up voice profiles and training it to recognize different voices through linking of accounts.

However, I wonder if there is a specific reason why the user experience of setting up voice recognition isn't as easy as other biometric information like finger print or facial recognition. Is this because the technology isn't as mature or that people don't use it as much as other types of biometric?

Are there good examples of design patterns for enabling better specificity of voice recognition where voice control is the dominant way of interacting with a system/application?