This dissertation presents a multimodal architecture for use in mobile scenarios such as shopping and navigation. It also analyses a wide range of feasible modality input combinations for these contexts. For this purpose, two interlinked demonstrators were designed for stand-alone use on mobile devices. Of particular importance was the design and implementation of a modality fusion module capable of combining input from a range of communication modes like speech, handwriting, and gesture. The implementation is able to account for confidence value biases arising within and between modalities and also provides a method for resolving semantically overlapped input. Tangible interaction with real-world objects and symmetric multimodality are two further themes addressed in this work. The work concludes with the results from two usability field studies that provide insight on user preference and modality intuition for different modality combinations, as well as user acceptance for anthropomorphized objects.
展开▼
