中文摘要：

循环特征检测法的检测性能远优于能量检测法,但是噪声不确定性会使这一优势变得不明显。针对这一问题,提出了一种能够对抗噪声不确定性的循环特征检测法。推导证明了在二维循环谱中,不同循环频率上的噪声分量具有独立同分布的特性。基于这一特性,利用授权用户信号循环谱峰以外的值来估计循环谱峰处的噪声分布,并获得不需要噪声功率先验知识的判决门限,从而避免了噪声不确定性的影响。实验结果表明,所提方法的检测性能接近噪声功率精确已知的循环特征检测法。

英文摘要：

The detection performance of cyclic-feature detection is much better than energy detection. However,noise uncertainty will degrade its performance severely. Aiming at this problem, a cyclic-feature detection method resisting to noise uncertainty was proposed. It was proved that the noise cyclic spectrum components at different cycle frequencies were independent and identically distributed in two-dimensional cyclic spectrum. Based on this,the noise distribution at the cycle peak was estimated using all values except the spectrum peak positions of licensed user signal, and decision threshold was obtained without noise power prior knowledge, so as to avoid the influence of noise uncertainties. Simulation results demonstrate that the performance of the proposed method is close to that of cyclic-feature detection method with known noise power.