A computation memory model with human memory features for autonomous virtual humans

Features of human memory are analyzed. A computation memory model following human memory features is proposed for autonomous virtual human. It is designed with three layers. The first layer saves top-N newly obtained or old but unforgettable knowledge. The middle layer saves indexes of top-N+1 to N+M knowledge that is no longer newly but easily recalled. The two layers are set in PC memory so that they are rapidly to access. The third layer includes a database for recording all obtained knowledge, and an accessorial database for exiting and restarting purposes. This layer is set at external hard disk. An algorithm is proposed to calculate knowledge retention in memory. In the algorithm, the Ebbinghaus forgetting law, the importance of knowledge, and its relationship with a person, are integrated to indicate the top-N+M knowledge in memory.