Multimodal speech interfaces for map-based applications

Abstract

This thesis presents the development of multimodal speech interfaces for mobile and vehicle systems. Multimodal interfaces have been shown to increase input efficiency in comparison with their purely speech or text-based counterparts. To date, much of the existing work has focused on desktop or large tablet-sized devices. The advent of the smartphone and its ability to handle both speech and touch inputs in combination with a screen display has created a compelling opportunity for deploying multimodal systems on smaller-sized devices. We introduce a multimodal user interface designed for mobile and vehicle devices, and system enhancements for a dynamically expandable point-of-interest database. The mobile system is evaluated using Amazon Mechanical Turk and the vehicle- based system is analyzed through in-lab usability studies. Our experiments show encouraging results for multimodal speech adoption.