The wireless signal of wireless local area network is unstable in the indoor environment, and the traditional support vector regression (SVR) based positioning method may lead to the reduction of the correlation between the position coordinates and signal strength. Thus, this paper proposes an improved support vector regression (ISVR) based indoor positioning method. Firstly, the logarithmic processing is conducted on the received signal strength (RSS) to make it more consistent with the normal distribution, and then the Gaussian filter is used to filter the small probability of fingerprints before building the fingerprint database. Secondly, in order to reduce the error of constructing X and Y coordinate model separately, a calibration coordinate z=x·y is trained at the training stage, which can improve the correlation relationship between RSS and XY position information. Finally, the optimal position coordinates are obtained by weighted inverse K nearest neighbor (WIKNN) method. The experimental results show that the proposed algorithm can reduce the noise caused by the complicated environment in the room, and has higher positioning accuracy than the traditional support vector regression algorithm.