It is important to select an effective subset of Acoustic Emission signal parameters to achieve the real-time requirements in wireless transmits. Currently, the parameters distance is wildly used in the subset abstraction. This method is easily disturbed by noise and impacting the resolution effect. In this paper, a new characteristic parameters selection method is proposed which combined the mutual information theory and distance measurement. The new approach not only considers the parameters distance, but also highlights the relevance between characteristic parameters and fault classes. At the same time, the redundancy of characteristic parameters is highly focused on. Experiments illustrated in this paper confirmed the improvement in parameters subset resolution and effectiveness using the new strategy.