As much as this is a great, unique interface mechanism, I think it highlights an important engineering principle: constraints.
The Nintendo DS is a portable system. You can carry it around in your pocket. You play it on a plane, train, or in an automobile. You might be playing it all alone on your couch or you could be in the middle of a crowd. How appropriate is voice control to this system?
It's not even so much that being loud is the problem -- unless you're in a library, church, or movie theatre, it's unlikely that making noise is a problem. The issue is with our expectations for social interaction and the cues triggered by somebody speaking.
As several anecdotes in this article reveal, we immediately respond to speaking as if somebody were speaking to us. Did you call me? Did you say to turn left? Who are you talking to? When something is being said, we expect it to be to somebody, and when it is said in the middle of silence, we immediately try to determine who is being spoken to, just in case it happens to be us.
It's the same etiquette problem that has developed (and is still evolving) with cell phones. The fact that somebody could be 2 feet from you, talking in a normal speaking volume, and yet not talking to anybody in the vicinty, is confusing to us. At least with a cell phone we can quickly ascertain that the person is talking on their phone; what goes through the mind of a person seeing you talk into a small black box?
As interesting and powerful as voice control is, it's not really appropriate to the platform. For something that will often be used in places where it's not appropriate to speak to a little black box, it doesn't make sense to require the use of a microphone. Thankfully some games (e.g. Brain Age) were developed with that in mind.