I'm sure it's possible, as when you get right down to it, programming is programming. Whether it's the best choice is another issue, as "best" will likely be dependent on the choice and quality of existing libraries that would implement the voice recognition/interpretation aspects of such an application.
