Speech-Based, Manual-Visual, and Multi-Modal Interaction with an In-Car Computer - Evaluation of a Pilot Study

摘要

This paper presents a pilot study comparing various modalities for control and presentation of a multi-modal in-car e-mail system. A simple interface for reading e-mail was constructed, which could be controlled manually by pressing keyboard buttons, by speech through a Wizard of Oz setup, or both. The e-mail program was presented visually on a VDU, read to the driver through speech synthesis, or both. Results indicate that in this context subjective task load was highest when manual/visual interaction was used. A solution may be interaction through user-determined modality selection, as results indicate that subjects judge their load lowest and performance and preference highest among the tested conditions when they are able to select the modality. Some evaluation issues for multi-modal interfaces are discussed.