Abstract
Purpose:
To test the viability of a voice-driven feedback system to assist low-vision subjects with mobility.
Methods:
The audible feedback system consists of audio bone-conduction headphones worn by the user behind the ear and a custom Android application, which delivers commands to the user when a researcher presses a button corresponding to that command on the program. Two low-vision subjects were recruited from the Braille Institute, Los Angeles. Wearing the audible feedback system, subjects were steered around a three minute randomized course five times by a researcher, who gave commands that were randomized by trial. Commands included ‘forward’, ‘veer left’, ‘turn left’, ‘veer right’, turn right’ and ‘stop’. The course measured 17ft X17ft and was interspersed with one foot cones every five feet. Appropriate response to commands, angle of turns and reaction time were measured. Subjects were also given an exit survey that measured the usability of the feedback system. The Android application recorded a time stamp for each command and video footage was recorded for each trial.
Results:
Subject one had an average of 71.80% responses consistent with commands, whereas subject two had an average of 53.18% responses consistent with commands; although subject two showed a higher range in improvement over the course of the trials. Both subjects had an average of 95% positive compliance to the ‘forward’ and ‘stop’ commands, and less than 5% compliance to the ‘veer left’ and ‘veer right’ commands. Subject one had an average reaction time of 1.5 seconds, whereas subject two had an average reaction time of 2.5 seconds. Both subjects had an average turning radius of approximately 70 degrees for the ‘turn right and ‘turn left’ commands. Based on the system usability scale, the system was determined to be 90% useable in its current condition.
Conclusions:
Judging by the reactions times and the system usability scale, the audible feedback system seems to be an intuitive solution that will create a balanced connection between the user and the feedback controller. However, significant training is needed to maximize appropriate responses by its users to the commands provided