Research on markerless visual based human motion tracking has been popular in recent years. This is because of the needs in applications, for example gait analysis, patient monitoring system, and surveillance system. This thesis proposed the development of markerless visual based human motion tracking using Kalman filtering algorithm and Sampling Importance Resampling (SIR) particle filtering algorithm. From the research results of this thesis, the processing speed of Kalman filtering is faster than that of SIR particle filtering. Besides, for human motion model that is more to linear and Gaussian, Kalman filter tends to yeild better accuracy. However, for human motion model that is more to non-linear and non-Gaussian, for example motion captured with a shaking camera, the SIR particle tends to yeild a highrer accuracy. This thesis also proposed a lower limb skeleton model detection algorithm. However, the accuracy of the proposed algorithm is prone to influence of noises and the subject's moving sytle.