The problem with using voice recognition to protect your computer is that a voice is both too lax and too strict.
Too lax:
It is actually REALLY easy, in this day and age, to record something spoken aloud. people can remotely activate webcam microphones and listen in on you, or they can "accidentally" leave their phone on your desk while at lunch, or they can even go the spy route and hide microphones in a variety of places, including ballpoints, your computer screen and even your overhead lights.
And even if you manage to avoid all that, text to speech technology has already progressed so far that you can generate a near-perfect copy of any voice. I just need to find a sample of you speaking and I can use that to generate a perfect or near-perfect rendition.
Too strict:
now, suppose that you find a way to make it listen to your speech pattern and only your speech pattern. What if your speech pattern changes for some reason? Voices change quite a bit depending on what emotions you experience, how awake you are and even your health.
And even if you can keep your voice pattern the same, there's still the problem of background noise. any office has background noise, be it the cars outside, the other people in the room or even the white noise from devices in your room.
Also, what if someone else has to access your computer in an emergency? Something like "You got into an accident on the way to work and a coworker needs a file on your computer RIGHT NOW!". If that happens, it's hard to tell your coworker how to access your computer because they don't just have to enter something in your keyboard, but they actually have to mimic your speech pattern.
Finally, while speech recognition is relatively easy (disregarding accents), voice recognition is actually tricky, since you can have 2 people with VERY similar voice patterns, but who need to be authenticated separately. You also need to safeguard against an attacker that has hired a professional impersonation comedian or voice actor to mimic your voice.
The crucial flaw for voice recognition based security is that you need a system that is flexible enough to not lock you out if your voice is slightly different, while being strict enough to prevent compromise through a simple recording or simulation.
I have no doubt that there are already people working on managing these problems, but I don't foresee a usable and affordable voice recognition authentication system appearing in the next few years.
