基于改进判决树的调制分类器

Modulation Classification Based on Improved Decision Tree Method

基于判决树的调制样式识别方法是一种实时性好、较为成熟的信号调制识别方法,但是如何正确确定判决门限是影响该方法稳健性的关键因素之一。论文提出了一种改进的判决树法分类器,利用神经网络的训练思想,对判决树法的参数门限进行训练选取,从而提高了判决树法的自适应能力,能够随着目标信号的变化而自适应调整判决门限。最后,通过仿真验证了论文所提方法的有效性。

Modulation classification based on the decision tree is comparatively mature and better in real-time for signal modulation identification,whereas how to determine the decision threshold correctly is one of the key factors influencing its robustness.An improved method of modulation classification based on the decision tree has been proposed in this paper. Using the neural network for reference,this proposed method selects the decision threshold by training,which improves the adaptive ability of the traditional decision tree. Thus,the decision threshold is adaptively adjusted along with the change of the target signal. Finally,the feasibility of this proposed method is verified by simulation.