Multimodal Human- Computer Interaction: a Constructive and Empirical Study Contents

Multimodal interaction is a way to make user interfaces natural and efficient with parallel and synergistic use of two or more input or output modalities. Two-handed interaction is a special case of multimodal interaction that makes use of both hands in a combined and coordinated manner. This dissertation gives a comprehensive survey on issues that are related to multimodal and twohanded interaction. Earlier work in human-computer interaction and related psychology is introduced within both these fields. The constructive part of this dissertation consists of designing and building a group of multimodal interaction techniques that were implemented in two research prototypes. The first prototype is an object-oriented drawing program that implements new tools that are controlled with two-handed input. The second prototype is a multimodal information kiosk that responds to both touch and speech input, and makes use of touch pressure sensing. In order to evaluate the success of constructive research, four empirical studies were conducted to compare the new interaction techniques to the conventional methods and to evaluate how the users react on them. The first of the studies compared a new direct manipulation tool to conventional menu and palette commands. The second evaluation was more informal and determined how an alternative way of drawing would be used by normal users. The third study was aimed at determining what is the best input device configuration to control the new tools. The last study evaluated different touch-based selection techniques in a multimodal touch and speech based information kiosk. The need for extensive interdisciplinary research is pointed out with a group of research questions that need to be answered to better understand multimodal human-computer interaction. The current knowledge only applies to a few special cases, and there is no unified modality theory of multimodal interaction that covers both input and output modalities. CR