To solve the problem of low accuracy and great noise, the gyro array technique is used to reduce noise and improve the performance of microelectromechanical system (MEMS) gyroscope. The measurement model is simplified on the basis of the component of noise in MEMS gyroscope, which is analyzed by Allan variance. Furthermore, a novel Kalman filter (KF) for combining outputs of a gyroscope array using cross correlation between noises from different gyroscopes is designed. The optimal estimation process is improved to decrease the complexity and calculation quantity, then the accuracy of gyro array and affecting factors are analyzed by using a steadystate covariance. Considering the dynamic performance of the signal, the true rate model is established by the autoregressive (AR) model. The experimental results indicate that the precision of the gyro array composed by six gyroscope increases 144.2 times in static condition. In dynamic condition, the precision increases 18.18 times when the input rate is constant and 5.36 times when input rate is sine function. It proves the validity of the modeling and fusion methods.