This paper presents a new calibration method to overcome the challenges of MEMS inertial sensors for underwater navigation. The MEMS inertial sensor module is composed of an accelerometer, gyroscope and circuit of signal process. For navigation estimation, it is easy to be influenced by errors which come from MEMS inertial sensors. In general, the sources of error can be categorized into two groups, deterministic and stochastic. The former are mainly including bias error, misalignment and nonlinearity; the latter contain temperature effect and signal drifting. Subsequently, the linearity calibration is used to modify the deterministic error and the wavelet analysis can suppress the stochastic noise. Therefore, the new calibration method integrated of linearity calibration and wavelet signal processing is utilized to enhance the accuracy and performance of MEMS inertial sensor module. The experimental results demonstrate that the output signal can be corrected suitability by means of proposed method.