Against the problems that indoor positioning suffers from low accuracy and poor stability, an indoor positioning method based on kernel function and Kalman filtering was presented.The Gaussian kernel function was adopted for initial positioning, which can measure the nonlinear similarity between reference points' fingerprint and test points' RSS(received signal strength), so it will get better accuracy and stability than the K nearest neighbor algorithm.Then the Kalman filtering was used for filtering the positioning results.The experimental results show that in the real WLAN(wireless local area networks)environment, the RMSE(root mean square error)of kernel function positioning results processed by Kalman filtering is decreased by 25%, the positioning accuracy within 2 meters is increased from 75% to 90%, and the positioning stability is increased by 29%.