I always chuckle at these, and yes voice independent recognition is always going to have this challenge. I tried a half dozen different "activation" phrases for my Moto-X and they all triggered at odd times by non-activation things (like the movies).

The really useful next step will be voice independent language recognition with voice dependent command recognition. That and accent independent language recognition. That is one of the, if not the, next billion dollar acquisition by one of the big players.