In this paper we present a novel IMU (Inertial Measurement Unit) based indoor localization system assisted by a fuzzy inference system (FIS). The positioning is achieved by simply applying double integration to the acceleration data. And the accumulative position errors are eliminated by using FIS based acceleration restriction. In this way, the system can be used for general cases without any scenario and condition limitations. By considering smartphone as the IMU, it is also possible to realize applications such as navigation in the supermarket or guiding blind people through buildings. Indoor experiment has been carried out to verify the benefit of FIS based acceleration restriction. The experimental results show that the proposed new system is able to follow the true trajectory and achieve a tracking accuracy improvement of 84% in comparison to the one without FIS based acceleration restriction.