The increasing need by the communities to understand every detail of events haveattracted the attention of researchers in machine vision and related work.This has led to anincrease in demand for smarter and cost effective video surveillance system which canaccurately distinguish between what is considered semantically normal or abnormal tohumans both in public and private places using intelligent vision.　　A skeletal tracking system that uses the Kinect sensor to monitor and record both theRGB and depth data is set up to study the behavior of individuals in public or private placesof interest.A posture or gesture recognition system is designed to monitor the individuals inthis human-computer interaction module.The recognition system is based on the templatematching dynamic time warping algorithm (DTW).　　This thesis discusses the design of a human computer interaction system that is able todetect any aggressive behaviors or postures such as pushing and the normal postures such as ahand wave.The DTW warps a time sequence of joint positions to reference time sequencesand thus generating a similarity value based on the recorded data which guarantees robust　　performance and allows computers to recognize any customized gestures by users.Also theXNA game 4.0 framework which allows the integration of body tracking into the DTWgesture recognition is introduced.