Human activity Modeling and Simulation (M&S) plays an important role in simulation-based training and Virtual Reality (VR). However, human activity M&S technology currently used in various simulation-based training tools and VR systems lacks sufficient biofidelity and thus is not able to describe and demonstrate the nuances of human activities and human signatures. This inadequacy becomes crucial when the training or the use of VR is human centered, such as human threat recognition training and dismount detection training. Human signatures that can be observed from a fairly long distance include body shape, gesture, and motion. In recent years, the Air Force Research Laboratory has investigated human modeling and simulation with high biofidelity, with an emphasis on true human shape and motion. This paper presents the technical development from these investigations, which include (a) static shape modeling and morphing; (b) pose modeling and dynamic modeling; (c) motion capture (in particular, markerless motion capture); (d) inverse kinematics and motion mapping/creation; and (e) creation and replication of human activity in 3-D space with true shape and motion. A brief review is conducted to discuss the methods and techniques related to these topics, along with some research results. Examples are provided to illustrate the importance of biofidelity in the simulation-based training.