Multimodal perception and recognition of humans with a mobile service robot

Abstract

Mobile service robots are becoming more and more popular, both in public and private places, so their perception capabilities must be adequate to detect and recognize people. One of the methods to accomplish the last task is face recognition, but this is unfortunately very challenging because of the human and robot motion. Also, besides the identities of individuals, the system should be able to distinguish between known and unknown people, and deal with this information accordingly. These challenging problems can be solved with a bank of Bayesian filters that simultaneously track and recognize the person of interest using laser and visual data. The paper extends this solution and proposes an improved version, which combines face recognition with human clothes and height identification, and that can also distinguish unknown people. The effectiveness of the system is demonstrated by several experiments with a mobile service robot.