A KPCA Detector Based on Observation Matrix of HLA in Uncertain Shallow-water Environment

Sound source detection is a difficult problem in uncertain shallow water. The horizontal linear array robust subspace detector (HRSD), which is a detector for horizontal linear array, requires little prior informations about uncertain enviromental parameters. But the detection performance of HASD degrades to be the same as the energy detector when the HRSD is difficult to further separate the signal from the noise in the subspace. We propose a detector based on kernel principal component analysis, which provides a higher detection performance with the same prior information requirement. The kernel matrix is constructed by the observation matrix of horizontal linear array. It is reasonable to construct the kernel matrix using the observation matrix when the array aperture is small. The simulation results show that the detector based on kernel principal component analysis has higher detection performance than HRSD under different environmental parameters. Meanwhile, the robustness of the proposed detector was close to that of the energy detector.