QuickFusion: Multimodal Fusion Without Time Thresholds

摘要

Multimodal user interfaces can provide natural and efficient interaction between humans and machines in a number of applications. Multimodal fusion integrates information from multiple input channels. Many multimodal fusion approaches exploit the temporal characteristics of inputs to determine the completion of a user's turn and cause a delay in the system response. This paper proposes a multimodal fusion approach, QuickFusion, which utilises syntactic rather than the time-stamp information from inputs to determine the integrity of the inputs, thus avoiding the time delay in the multimodal fusion process. QuickFusion also helps to resolve input ambiguity.