In recent years, head-mounted displays (HMDs) have become
increasingly affordable and popular. HMDs pose new
challenges for text entry because they typically immerse
people in a virtual world while preventing them from seeing
the real world. As HMDs become widespread, we expect
that some people will want to perform traditional computing
tasks with these new devices, such as writing emails or
browsing the web. This work explores new ways to support
efficient text entry in HMDs by leveraging decoding techniques
and ubiquitous physical keyboards. We describe
an experiment where we measured the performance of
our system on touch-typists who typed a series of short
messages on a physical keyboard in a (1) keyboard visible
condition, (2) occluded keyboard condition, and (3) headmounted
display condition. The results of our pilot study
showed that users’ speed and accuracy was considerably
worse in the occluded and HMD conditions, that our decoder
was able to correct a large number of errors, and
that increasing compute time for the decoder increased the
number of corrected errors.