Automatic face recognition system architecture for collaborative virtual environments

In a collaborative virtual environment (CVE) participants are represented by icons called avatars. In order to correctly assign these avatars, participants must be identified as they enter, leave and navigate within the CVE. We present an approach and systems architecture for the use of automatic face recognition (AFR) to accomplish the task of participant identification. The requirements for AFR within a CVE are discussed and distinguished from those for general applications of AFR. A face recognition server has been developed to implement these requirements, and is described. The system is implemented in a Java Server Pages environment, using Java Native Interface to interact with individual face recognition algorithms, and interacts with other CVE components using messages in the SOAP protocol format.