The Study of Two Novel Speech-Based Selection Techniques in Voice-User Interfaces

The voice-based technology is a promising alternative input modality to the conventional computer input devices, such as keyboards and mice. For mobility-impaired people who cannot use the traditional computer input devices, a voice-based interactive solution is highly desirable. Although some grid-based cursor control systems that use speech recognition show potentials, they come with limitations in processing time and difficulties in real-time applications. Since they are not natural enough to use, it is very slow to drill down the grid, and the point cursor selection range is prone to errors, and people need a more efficient and natural choice. To address these limitations, this paper proposes two novel speech-based target selection techniques: grid-based region selection (hereinafter referred to as “Region Selection”) and grid-based target search (hereinafter referred to as “Target Search”). Region Selection allows users to select targets directly using the selected area. Target Search permits users to move the cursor to a target in a specified direction. Experiments are designed and conducted to analyze the impacts of target size and target density on performance and processing time. The experimental results and comparison study with traditional solution show that both Region Selection and Target Search improve the participants’ accuracy of target selection and reduce the processing time, especially for small targets.