Mars Papers

McGowan, JF (2005).
Talking To A Computer On Mars
Preprint.

Abstract:
Astronauts exploring Mars will probably want to use computers in the field for many purposes including communication such as e-mail, information retrieval, and command and control of both their spacesuit and off-suit equipment. Traditional computer systems present many difficulties in the field including the difficulty of operating a keyboard wearing a spacesuit and the effect of pervasive dust and extreme conditions on equipment. Speech-recognition and speech synthesis have been proposed as partial solutions for these problems.

Design studies and experiments using an analog spacesuit simulating the use of speech recognition for communication, command and control by astronauts on Mars are presented. The studies and experiments were performed by the Northern California Chapter of the Mars Society and GFT Group Inc. using state-of-the-art commercial speech-recognition products, Nuance Communications Inc. (formerly ScanSoft Corporation)'s Dragon NaturallySpeaking Professional and GFT Group's artificial personal assistant Petrana. Many of the results will be applicable to astronauts on the Moon, other planetary bodies, and in orbit.

Other related work on speech-recognition for astronauts and spacesuits is reviewed briefly. Methods for creating a controlled acoustic environment in a spacesuit so that current large vocabulary, speaker-dependent speech-recognition can be used reliably are discussed. Display options such as in-helmet displays, external projectors, and external flat screens are discussed. The advantages and challenges of in-helmet displays are detailed. Pointing devices such as trackballs and motion sensors that can be integrated into a spacesuit are discussed for operating graphical applications and traditional graphical user interfaces. User interface and operating system design with no keyboard, no mouse, and no traditional graphical user interface is discussed.

The size, weight, and power requirements for a fully speech enabled system are calculated. In-suit and off-suit speech processing are compared. Power supply options such as a battery, fuel-cell, or solar cell are discussed. The integration of speech control with sensors such as a video camera, trace gas monitors, and so forth is discussed. Partitioning applications between fast, high reliability physical input devices -- switches, buttons, motion and pointing sensors -- and slower, less reliable, more flexible and versatile speech input is discussed. In general, mission-critical and life-support functions will use physical inputs. Lower priority and support functions such as e-mail, database query, and non-essential equipment control will use speech. Strategies to avoid problems or disasters from speech-recognition errors, mispronunciations, and so forth are discussed.

What can and cannot be done in space exploration applications with current accuracy large vocabulary speech-recognition (about 95% accurate) is discussed. Some strategies to increase accuracy such as incorporating syntactic and semantic rules are discussed.

PDF summary of all papers (not books or other publication types)