The data fusion algorithm has been developing extensively with the wide application of sensor networks. An Informax-based data fusion algorithm is proposed in this paper. The fusion matrix can be self-adaptive from the mutual information of the sensors output to obtain the maximal approximation of the original signal. In addition, the effects of step-size factor are analyzed. A Gyroscopes Network that uses the Informax algorithm to fuse the gyroscopes output is introduced as an implementation. Furthermore, a prototype has been built and the precision of the gyroscopes network after data fusion was improved about 6.5 times over that of a single gyroscope in prototype.