Existing localization algorithms which use Kalman filter for indoor environment set Gaussian distributed noise covariance, or change a noise covariance by complicated filters. If noise covariance is considered as Gaussian distribution, the localization system can not reflect real environment, so the Gaussian distributed noise covariance makes localization system unpractical. Also complicated filters limit the applications which localization algorithms can apply. In this thesis, we propose a method that adaptively changes noise covariance. The proposed method recognizes changes of moving pattern based on results of localization and previous state, and then adjusts the noise covariance. In addition, the method varies localization period to catch up rapidly with mobile nodes``s movement and reduce overhead caused by calculation of location when mobile node moves constant pattern. To verify the performance of proposed algorithm, we simulated in various situations. Simulation result from MATLAB shows the proposed method outperformed an algorithm which has constant localization period and Gaussian distributed noise covariance of Kalman filter by 8%.