In order to decrease temperature related errors in MEMS gyroscope,a novel method based on array technique is proposed for temperature compensation of the gyroscope.The advantage of this method is that it is not need to measure the temperature,in stead,data fusion of the output of several gyroscopes is used to reduce the influence of temperature,and the system structure is simplified.The BP neural network optimized by genetic algorithms is used in data fusion of the gyroscope data,which compensates the error of the output of gyroscopes directly,while the zero bias and scale factors are compensated separately in general.The results of the experiments show that the method proposed can significantly reduce the error of gyroscope caused by temperature,and thus can improve the accuracy.