基于矩阵重构的信源数估计方法

Signal Number Estimation Based on Reconstructed Matrix

提出了一种基于矩阵重构和多重假设检验相结合的方法。采用特殊定义的变换-对消矩阵,抵消了非平稳噪声的影响;采用多次重构的显著性函数结果与显著性水平比较来判断信源数,避免了假设检验对先验知识的要求。将单序列假设检验扩展为多个序列联合检验后,即使在低信噪比和低快拍数时也可以获得较好的估计性能;同时显著性水平的确定与阵列数据无关,基本不需要先验信息,约束条件较少。仿真结果验证了该方法的有效性。

Based on the reconstructed matrix and the multiple hypothesis testing, this paper presents a scheme for estimating the signal number in spatial nonstationary noise fields. A sp- ecial transform-differencing matrix is constructed to cancel the nonstationary noise. Then, the result of multiple reconstructed hypothesis tests is used to estimate the signal number compar- ed with the significant level, thus avoiding the requirement for prior knowledge of hypotheses. The scheme has a superior performance even in the case of low SNR and short snapshots. Furt- hermore, the significant level has no a relation to the receiving data of the array, so it does not need any priori knowledge. Simulation results prove the improved performance of the method.