On the way forward to real time in the process of particle-filtering-based human hand tracking, one of the main obstacles is to generate a great deal of particles which are derived from high dimensionality of human hand model. Aimed at reducing the particle number, a new particle filtering approach is put forward in this paper. First, the operator's cognitive psychology features in the process of human-computer interaction are analyzed and studied and a general dynamic motion model is constituted. Second, some basic features of the dynamic motion model are studied and mathematically described. Third, a zonetime model of the moving human hand is proposed, and furthermore, the features in time-space of a hand gesture state are discussed. Based upon the abovementioned job, a new concept, microstructure of state variable, is presented, upon which the novel hand gesture tracking algorithm is put forward. Finally, experiments are implemented including some comparison experiments, and the algorithm is also compared with some referenced algorithms. The main contribution is that the study describes and models hand gesture behaviors and connect them with freehand tracking. The experimental results show that just using a small quantity of particles, compared with the referenced algorithms, the algorithm can obtain satisfactory results.