Smartphones are built with a wealth of sensors, which are characterized by small size, lightweight and low cost, etc. With the popularity of smartphones, developments and applications using smartphone's built-in sensors have attracted increasingly research interest in recent years. However, low-cost sensors have inherent disadvantages, such as low measurement accuracy and poor measurement stability. Therefore, in practical applications, the errors of smartphone's built-in sensors cannot be simply ignored. To estimate the sensor errors, this paper analyzes the error characteristics of low-cost accelerometer, gyroscope and magnetometer respectively, and calibrates each sensor by using improved six-position method, Allan variance method and ellipsoid fitting method. Experimental results of sensor error estimation demonstrate the effectiveness of the calibration methods on the built-in micro-electro-mechanical system (MEMS) sensors of smartphones.