A couple of weeks ago our ECE department hosted Bret Harsham of Mitsubishi Electric Research Labs (MERL) as a part of the ECE900 graduate seminar. Bret gave a very interesting lecture about the ways of shortening voice dialogs through using a contextual push-to-talk button. The title of the lecture was “Contextual Push-to-Talk: Shortening Voice Dialogs to Improve Driving Performance.”
The focus of the lecture was on a prototype in-car voice user interface (VUI) which was tested during my MERL internship last summer. As opposed to the contemporary in-car VUIs, which use only one push-to-talk button for issuing commands, this work presents a way of utilizing multiple push-to-talk buttons depending on the context of the query. For example, if we have three domains of interest, we can associate one push to talk button for each of them. Therefore, we skip multiple steps which are otherwise required in order to switch to a desired domain and initiate a voice search.
This work was recently accepted for publication at MobileHCI 2010 conference.