频谱预测是一种通过分析历史频谱测量数据获得频谱使用规律,从而预测未来频谱使用状态的技术。在认知无线电中,认知设备根据此频谱预测信息,可以实现智能的频谱感知和动态的频谱接入,降低开销,提高通信质量。文章首先对现有频谱预测方法进行了分类,然后依次介绍了各种频谱预测方法的基本工作原理,在此基础上对比分析了各种方法优缺点,最后探讨了存在的技术挑战和下一步研究方向。 Spectrum prediction is a technique that obtains the spectrum usage pattern by analyzing historical spectrum measurement data to predict the future state of spectrum usage. In cognitive radio, cognitive equipment can realize smart spectrum sensing and dynamic spectrum access according to the spectrum prediction information, reduce overhead and improve communication quality. The article first classifies the existing spectrum prediction methods, and then introduces the basic working principles of various spectrum prediction methods in turn. Based on this, the advantages and disadvantages of various methods are compared and analyzed, and the existing technical challenges and the next research are discussed direction.