In a pen-based computing system, a pair of microphones captures audio concurrently with a smart pen device capturing handwriting gestures. The audio captured by the pair of microphones is processed to produce an enhanced audio recording. Noise cancellation techniques may be applied to reduce noise generated by the smart pen device interacting with the writing surface to produce a higher quality audio recording. In addition, beam steering techniques may be applied to adjust the gain of portions of the audio signal from different directions and allow the user to focus between multiple audio sources.