It should be a lot easier for a microphone that's right next to your head. Face one or more microphones away from the user, that gives you the environmental noise. Point one or more microphones at the user's mouth (or use some other method). That gives you what you need (signal much louder than the environmental noise, with the right microphone), along with some of that same noise. Apply some noise subtraction magic.

It's actually pretty different than an Echo, which needs to pick out the user's commands via complex signal processing alone.