I've got voice recognition in my computer. It's an optional feature built into the OS, which you can turn on. But you have to press a key (that you choose) prior to speaking a phrase. It was novel for a couple of weeks but soon became annoying. "Logoff", etc. spoken commands just seemed idiotic after a while. Sort of like "clap on" for turning on lights.
Voice recognition is awesome for movies and TV because they turn it into dialog. In practice though it’s a slow error prone way to interact with things.
That is a much more sensible implementation of any voice command. . . invoke it's attention with a button push. . . or an on device pass phrase the device is capable of parsing before the server is connected to interpret more complex instructions or requests similar to the way it is handled by Apple with Siri or Microsoft with Cortana.