Abstract:

The special spectral estimation using Capon approach can be achieved by measuring the generalized inner product (GIP) value between each search steering vector and the correlation matrix (CM). Therefore, the CM deviation, which is induced by the secondary data number is comparable with the sensor number and the power of one signal is much stronger than that of the others, will deteriorate the performance of Capon algorithm. To alleviate the performance degradation, a new modified Capon method is presented. Firstly, the CM is estimated with iterative approach which effectively relieves the requirement of a large number of training samples. Secondly, a modified correlation matrix at each search steering vector to calculate the GIP value is employed. Finally, the spatial spectrum can be obtained as weighted Capon spectrum with the GIP value. The high resolution property of Capon approach is remained, meanwhile the robustness against small sample support and strong and weak signal coexistence is also enhanced. Theoretical analysis and numerical simulation indicate that its performance is better than that of Capon.