Spokesperson detection Sample Clauses

Spokesperson detection. Using nonverbal cues provided by the human detection modules and other social processing modules, the spokesperson module will identify a person in a group that can be approached and addressed. What it expects: • An array of detected groups (provided by the group detection and tracking), each with – A unique group track id. – The number of persons in the group. – An array of the members of the group (identified by unique person id, see below). – Position (x, y, z) of the center of gravity of the group (preferable at 30 Hz. or more). • An array of detected persons (provided by the people detection and tracking), each with – A unique person id. – Position (x, y, z; preferable at 30 Hz. or more in world-coordinates). – Velocity (x, y, z; preferable at 30 Hz. or more in world-coordinates). – Length of the current motion track. – Human attributes (gender, age, height) and if possible the type of traveler (i.e., business traveler, family traveler, ...). – Positions of the upper-body joints (x, y, z; preferable at 30 Hz. or more in world-coordinates). – Length of the current upper-body track. – Head pose (angles in world-coordinates; preferably at 30 Hz. or more). – Length of the current head pose track. – A classification of who looks at whom for all persons in each group (e.g., for three persons this would be a 3x3 array with each entry xij between 0-1 indicating the likelihood that person i is looking in the direction of j person).