A camera-joystick for sound-augmented non-visual navigation and target acquisition: a case study

This paper presents the results of a comparative study of user input with a camera-joystick and a manual joystick used in a target acquisition task when neither targets nor pointer could be perceived visually. The camera-joystick is an input technique in which each on-screen item is accessible from the center with a predefined vector of head motion. Absolute pointing was implemented with an acceleration factor of 1.7 and a moving average on 5 detected head positions. The underlying assumption was that, in order to provide a robust input for blind users, the interaction technique has to be based on perceptually well-discriminated human movements, which compose a basic framework of an accessible virtual workspace demanding minimum external auxiliary cues. The target spots, having a diameter of 35 mm and a distance between the centers of adjacent spots of 60 mm, were arranged in a rectangular grid of 5 rows by 5 columns. The targets were captured from a distance of 600 mm. The results have shown that the camera input is a promising technique for non-visual human–computer interaction. The subjects demonstrated, more than twice, better performance in the target acquisition task with the camera-joystick versus the manual joystick. All the participants reported that the camera-joystick was a robust and preferable input technique when visual information was not available. Blind interaction techniques could be significantly further improved allowing a user-dependent activation of the navigational cues to better coordinate feedbacks with exploratory behavior.