Interaction Region Characteristics for Midair Barehand Targeting on a Television

Recent computer vision advances can enable television control using midair barehand input, for which selecting targets using hand positions is a primary task. Yet, the size and position of a preferred 2D interaction region has not been specifically investigated for this setting. We report on a field study with people in front of their own television in their own home. Controlled variations of target position stimuli were presented while a camera records the natural hand position used by the participant in response. Based on hand and face landmarks, density plots define a preferred input region location and size using a human-scaled unit of face-widths. Distribution of hand positions relative to target stimuli reveals consistency and precision, suggesting an ability for users to map their input space to display space. We elicit an ideal input region from participants and provide a range of dimensions for designers implementing barehand input techniques.