A new method of spatial-spectrum detection based on noise suppression in signal subspace

In this paper, a new method to overcome the poor spatial-spectrum detective performance of Minimum Variance Distortionless Response (MVDR) is proposed for weak broadband signal. The new approach, called EBMVDR in this paper, is based on noise suppression in signal subspace. Firstly, Eigen decomposition is used in covariance matrix of the received data vector in frequency domain, and then estimate noise power using the pre-estimated signal dimension. Next, the covariance matrix is reconstructed by suppressing noise in the signal subspace. Finally, the weight vector can be got for the new method. At the same time, array gain of EBMVDR is deduced. Theory analysis and statistical simulation results show that: when the noise is white, the lowest detection signal-noise ratio (SNR) performance improves 6dB compared to the conventional MVDR method. Moreover, the new method has good robustness and stable array gain within the whole bandwidth.