The microphone and the Camera have in common that recording, storage and playback are already useful (in fact the same holds for the pen, too). Because audio has lower demands than video, much more is possible. Some application could be controlled by speech recognition. Or the incoming sound itself could be processed in some way (filtered).
The most useful application of sound in computers will be speech recognition. Several methods already proved to be capable of doing reasonable recognition in real-time. Especially the Hidden Markov Model (HMM) is becoming popular.
Esprit Project 8579/MIAMI (Schomaker et al., '95)